I studied in depth how to be successful in my PhD applications and it paid off: I got admitted to Stanford, University of Washington, UCL, CMU, and NYU. This blog post is a mish-mash of how to proceed in your PhD applications from A to Z. It discusses what is important and what is not. It discusses application materials like the statement of purpose (SoP) and how to make sense of these application materials.
There are some excellent sources out there on this topic and it is worth stopping for a second and understand what this blog post will give you and what other sources can give you. This blog post is mainly focused on PhD applications for deep learning and related fields like natural language processing, computer vision, reinforcement learning, and other sub-fields of deep learning. This blog post assumes that you already have a relatively strong profile, meaning you probably have already one or multiple publications under your belt and you worked with more than one person on research. This blog post is designed to help you optimize your chance for success for top programs.
If you seek more general information for PhD admissions, I recommend reading all the most highly voted questions and answers from Academia StackExchange. Other important sources are Applying to Ph.D. Programs in Computer Science which is a detailed write-up of the full admission process as viewed by CMU faculty. A similar but more concise source — in particular, relevant for good but not strong candidates — is the blog post Reflecting on CS Graduate Admissions which is again by CMU faculty. Less useful, but a quick read is the negative view of How to Write a Bad Statement for a Computer Science Ph.D. Admissions Application.
This blog post will first define what is important in PhD applications. Then we will dive into the application materials and how to think about these. Then I will talk a bit about the application process. The final section of the main part of this blog post will be on selecting schools — which schools are too good or too bad for me? After that, I will close with a Q&A section which was drawn from questions on Twitter. I will update this Q&A section periodically. If you have some questions regarding the application process, please leave a comment and I try to get back to you.
Understanding What Makes a Strong PhD Application
The most important factor that determines admission at any research university is research potential: How likely are you to become a great researcher? The main direct indicators for this are in order of importance:
- Recommendations: Respected professors speak highly of you. Personal connections are important.
- Research experience: You did successful research before. Measured in publications, first-authorship, and prestige of conference where you published.
Other indirect factors can help sometimes if they are exceptional, but usually, only the first two factors, recommendations, and research experience count. In order of importance:
- Undergraduate university name: Some universities select aggressively for this, some others do not care so much.
- Employer name: It is common that students are admitted that were previously employed in finance or at companies such as Google, Facebook, etcetera.
- Smarts: Perfect GPA, perfect GRE is somewhat correlated with intelligence (or at least with how fast you can learn and understand).
- Grit / Conscientiousness: You do well under continuous rejection, disappointment, and failure. If you faced and have overcome difficulties you might want to include your story in the statement of purpose.
- Accomplishment: You won Math or CS competitions.
- Recognition: You won prestigious scholarships/fellowships.
- Good at math or engineering: You developed or contributed to open source projects. You worked with research code.
- Heritage: Parents are professors.
Understanding Application Materials
Understanding Recommendation Letters
For recommendation letters, one could devise four categories: Strong, Good, Weak, and Bad. Note that the main thing that admission committees look for in recommendation letters are indicators of research potential. This section has the main purpose of making you aware of what constitutes a good or strong letter and based on this information it might be easier for you to select letter writers.
Signs of a Bad Recommendation Letter
- Your letter writer knows you and writes bad things about you. Especially in the US anything even slightly critical is very bad.
- Your letter writer does not know you (you had a class with her but you left no impression).
- Your letter is short and only states that you did well in class.
Signs of a Weak Recommendation Letter
- Your letter writer knows you from class only.
- Your letter writer is favorable, but can only write about achievements in class: Great project work in class; part of lively, interesting discussions in class.
- The letter writer does not comment on your research.
- The letter writer is not known by the admission committee nor by potential advisors.
Signs of a Good Recommendation Letter
- The name of the letter writer is known by parts of the admission committee.
- The letter writer’s name and work are known by at least one potential advisor mentioned in the statement of purpose.
- The letter writer worked with you on research.
- The letter writer mentions your excellent research abilities in anecdotes that demonstrate your creativity, commitment, persistence and research skills in general.
- The letter writer writes about how you published your research.
- The letter writer comments about research done outside of her lab.
Signs of a Strong Recommendation Letter
- US-style recommendation letter: The achievements are oozing through the paper. Everything is very much overdone, that is simple things become grand achievements.
- The letter writer has an excellent command of English.
- The letter writer is personally known by at least one potential advisor mentioned in the statement of purpose.
- The letter writer is known for making excellent recommendations (previously recommended students do very well).
- The letter writer mentions your excellent research abilities in anecdotes that demonstrate your creativity, commitment, persistence and research skills in general.
- The letter writer mentions your abilities which help indirectly with research (engineering skills, presentation skills, interpersonal skills) and wraps these skills into anecdotes.
- The letter writer comments about research done outside of her lab.
Note a few things:
- Anecdotes are important because the show that the letter writer really knows you. They also read much better. Stories are more interesting than checklists.
- The letter does not need to contain everything listed here to be considered “bad” or “strong” and so forth. Recommendation letters are complicated.
- If you select recommendation letters it can make sense to have some diversity among letters that highlight different strengths. One strong letter on research skills, a good letter on engineering skills (internship), and a good letter on performance in class/project work is a great combination. This combination is better than a strong letter on research, a good letter on research, and a weak letter on research.
- Please see more details about the process of asking about recommendation letters below.
Understanding Publications
Author Position
Publications are direct evidence for research experience and research skill. If you published as a first author, people know that you did most of the work. If you published as a second author, people know that you did a good portion of the work (25%-50%). If your name is the third or later, your contribution is discounted, but you probably went through the entire research process towards publication and gained a good amount of research experience. If you published a couple of first author papers a third author paper looks very good: It shows that you can work in a team.
Prestige of Venue
If you published your work at a respectable conference, people know that: (1) Your work is high quality; (2) your work can be trusted; (3) that your current research skills is sufficient to publish at great conferences, (4) that you are competitive and/or you can stay productive under the pressure of publishing at a top conference.
It helps to view this in the eyes of a potential advisor: If you have two students, one published already at NeurIPS (Tier A) and one you published at a Tier B conference. You would know that the first student is probably ready to work on a research project which is aiming for NeurIPS next year. The second student would need further preparation, for example, publish in a workshop or at a less competitive Tier A conference before making the step towards NeurIPS. With the second student, there is some risk that this student might take more than a year to acquire the research skills to needed to publish at Tier A conferences. Pushing a student towards NeurIPS can be stressful for an advisor and it is easier to work with someone who already has the necessary research skills. If there is less stress between advisor and student then its easier to develop a strong professional relationship which makes it easier and fun to work with each other. So a potential advisor would have good reasons to select according to the prestige of the conference where you published at.
Creativity, Citations, etcetera
Other indicators have little effect on the application. Your work might be unusually creative, but you have no track record that you are a creative researcher. Maybe you got lucky.
The importance of publications often only emerges with the years. Often you published shortly before the PhD applications which means that the citations that you have on your work is a poor indicator of impact. If you get an usually high number of citations in a short time this can help, but maybe you just got lucky or good at marketing. Usually, the number of citations over the past 1-3 years is no reliable indicator of research potential and as such is disregarded. If you have a citation history over the past 5 years this might be a different story, but this does not apply to most applicants.
Understanding the Statement of Purpose
For most institutions, the statement of purpose is mainly a filter for people who took no time to polish the SoP. Your writing can show how you think, how you can sell, how you explain things, but it can also show that you are lazy and do not pay attention to details. It can show that you are not able to Google simple recipes of how to write (and how not to write) a simple formal document. For some institutions, the SoP can be important (CMU) but the content does not really differ for these institutions.
Beyond formalities, the SoP is also the only document where you can justify why you did underperform in certain circumstances. For example, you can explain any extraordinary difficulties that you had along the way to graduate school, or it can explain why you did not do so well in certain semesters/quarters at uni. The structure of a SoP should be the following:
- Intro to research interests with a catchy hook that makes the reader want to read more (one paragraph). This is the most important bit: If you do not interest your readers in this paragraph it is unlikely that they will focus on the rest of the letter.
- The research experiences that you gathered along your way to grad school (about one page).
- Identifying what research you want to do in the future.
- Identify people with whom you want to work with and why.
- (Optional) Explaining extenuating circumstances where appropriate.
In some circumstances, the SoP can be very important. This is so if you showed good — but not strong or weak — academic potential and you had to overcome significant hardship to be able to do research. If your application is strong and write about hardships it might alienate your readers (privileged prick); if your application is weak it might also alienate your reader (whining looser). If your application is good it is exactly right (a smart person that pushed through difficulties). For example, I had a rare situation where I was barred from university access and my SoP was very important to explain the difficulties that I faced under these circumstances.
However, disclosing hardships and weaknesses — like learning disabilities and mental illnesses — can also be double-edged sword: You might either alienate the readers or you might draw their sympathizes and admiration for persisting in a difficult situation. If you disclose such facts, it needs to be done right and the SoP needs to be extremely polished. Do not attempt this if you do not have the feedback from expert writers. For some stories which are more socially acceptable you do not need expert feedback to do it right: It is easy to write a compelling story where you worked yourself from extreme poverty into college and that you now want to realize your potential by doing a PhD; it is difficult to write a compelling story about the hardships that you faced while suffering from schizophrenia or bipolar disorder.
However, if you did not face any hardship do not make up stories that make no sense: “As a white, male, upper-class US citizen, I was haunted by the responsibility of my privilege from an early age and my academic performance suffered in the process.”, instead, concentrate on your research experience.
Understanding GRE, TOEFL, GPA
The GRE & TOEFL tests and GPA are usually used as filter criteria. A very high GPA can be a good indicator of “some intelligence” which can help with the recommendation letters and publications are borderline. But a GPA of 4.0 will not help if you have no publications and bad recommendation letters — it might even hurt you because it shows that you concentrate on useless classes rather than research. GRE and TOEFL scores are pure filters: If you have an okay score you are not filtered out. If you have a perfect GRE score, it can help a little bit but much less so than a perfect GPA. Great GRE scores do not matter: I got into three out of the top five US computer science programs with verbal 159 (81%), quantitative 163 (86%), writing 5.0 (93%) and a TOEFL 120/120 and a GPA of 8.1/10. Any GPA higher than 3.5 is good. Anything above 3.5 does not matter. A GPA of 4.0 might help a little bit.
Understanding the CV
The CV lists what you have done. There are no surprises here. The content is important but the content is also determined by what you have done before and cannot be changed. Do not “tune” your CV by phrasing things in a nice way or by making your CV look “nice” or “creative” — this is a waste of time. Just list what you have done.
The Application Process
How to ask your professor for a recommendation letter
You write two emails: (1) Just ask if the person can write you a good or strong recommendation letter. Knowledgeable recommendation letter writers will reject your request if they think they cannot write you a good letter. In this case, look for someone else. (2) If your recommender agrees she will ask you to include some information for the letter. Give a list of what you have done with the person. Write it in a style that can be easily wrapped into anecdotes:
- DO: “You told me in a meeting that with some extra work we could make it for the NeurIPS deadline. In the next two weeks, I develop an improved deep network architecture started writing up the findings. The next week, Jane extended my code for an additional task. We then had enough results to submit our work to NeurIPS”
- DO not write: “Jane and I published our research at NeurIPS.”
Anecdotes can also come from interactions with PhD students and post-docs:
- “I worked with Tom on developing the research library that served as the main framework for our research that we published at NeurIPS. I worked one week on the library and Tom told me that the library was well designed and well performing.”
Your advisor will then ask the respective PhD student or post-doc for more information to write something like this:
- “My PhD student Tom — whom I regard as one of my most engineering-savvy students — worked with Jane on a research project where we needed to develop a code-base for language modeling before we could start the research. Tom gave this task to Jane and estimated it to take 3 weeks. Jane completed it within one week. Tom told me that after he inspected Jane’s code in a code-review, he found that Jane’s engineering abilities are on-par or even exceed his own — the code was very high quality and lightning fast. Jane’s engineering skills helped with the rapid development of research ideas. The research project became a walk in the park because of this. Jane published her work at NeurIPS2020…”
(2) If you have three letters which are on or above the “Good” level, you should think about making your letters more diverse. I for example used one academic letter, one industry lab letter, and one letter from a lecturer who is aware of my research.
Statement of Purpose
Start early and ask experienced people for feedback. You should be safe if you follow the formula above. If you want to disclose difficulties that you had along the way to graduate school you will need a lot of time in your SoP and you can expect that the SoP will take by far the most time in all your application materials.
Try to reuse letters between universities. It takes too much time to “personalize” the SoP for universities. The only section that I changed in my SoP from university to university was the section that mentions the potential advisors I would like to work with.
Online Application
Start early filling out the online applications early. Some forms are terrible and take some time to fill out and it is great if you can get this out of the way as early as possible to focus on recommendation letters, university selection and your statement of purpose. You should have a good reserve of money to do these applications. The entire process might cost up to $1000. If you do not have the money, ask some relatives for some help early on.
How to Select Schools for PhD Applications?
Can I get admitted to a top school?
Many people reading this probably have the dream to get into a top school like Stanford, MIT, Berkeley, or CMU. But admission is really tough. Some programs are highly selective. Here admission statistics for one top school I was admitted to and the prior probability of getting admitted to the program. Note that I have hard statistics on the schools and publications, but I do not have hard statistics on the letters and personal connections but I make assumptions based on what I have heard and seen from admitted students that I talked to:
- Top 2 undergraduate school AND 1 to 3 publications AND >=1 strong letter AND personal connections: 38%
- Top 4 undergraduate school AND 1 to 3 publications AND >=1 strong letter AND personal connections: 14%
- Top 20 undergraduate school AND 2 to 4 publications AND >=1 strong letter AND personal connections: 21%
- Below top 20 undergraduate school AND best school in a country (Tokyo, Australian National) AND 2 to 4 publications AND 1>= strong letter AND personal connections: 11%
- Master in top 3 school AND 1 to 4 publications AND >=1 strong letter AND personal connections: 5%
- Below top 20 undergraduate school AND not the best school in a country AND >4 publications and >=2 strong letters AND personal connections: 5%
- Below top 20 undergraduate school AND not the best school in a country AND >3 publications and >=2 strong letters AND award for Best Teacher/Young Scientist AND personal connections: 5%
This program, like most top programs, selects aggressively for undergrad degree. Note that usually, some form of personal connection (a letter writer knows a possible advisor at the school) is a requirement especially for edge cases. Other top programs select differently. For example, while CMU also selects aggressively for undergrad degree, they also like candidates with an unusual background which reflects strong performance under difficult circumstances. Some schools really like awards in math/CS competitions. Many schools like it if you got some form of best teacher award. Some schools like it if you have a portfolio of hacks (MIT). However, in general, in order of importance to get admitted to top schools:
- Personal connections
- Top undergrad school AND publications
- Strong letters AND publications
- Publications
- Anything else
This means if you doing an undergrad at a top 2 school and you have no publications you will still have a hard time. Top 2 school and a publication increase your chances of admittance dramatically. If you have no personal connections it is difficult to get admitted even with a strong profile. However, if your profile is overly strong under respected advisors then personal connections do not matter.
There are some other factors for special cases. For example, if you study at a top school and have only 1 publication then GPA will be an important factor. However, in general, top schools do not care about GPA numbers from schools below top 20 if it is at least a GPA of 3.5 or equivalent. So if you have a GPA of 3.5 at a below top 20 school and you have 4 publications you have a good chance of getting admitted. A low GPA (which is still > 3.5) can be a factor in favor if your research profile is very strong as it demonstrates that you do not care about classes but that you are passionate about research — exactly what advisors want to see.
Another thing to note here is that we have publication inflation. This means the value of a single publication becomes less and less because more and more students fulfill this requirement. The more students are interested in ML PhDs the more stringent the publication requirements. It might have been fine to have no publications to get into an ML PhD, but this is often no longer the case.
How to get admitted to top schools?
These statistics above do not mean that you cannot get accepted by these schools, but it means that if your profile is too weak you should take another year to bolster it. I, for example, extended my master by a year to squeeze in a year of research internships. Without this, I would never have made it into these schools. If your dream is to get into one of these top schools this is by far the best option. Even if you do not necessarily want to get into top schools, a research internship is highly recommendable.
A research internship will give you:
- Improved research skills so you can get an easier start into a PhD.
- A test whether a PhD or a certain research direction (NLP vs computer vision vs systems) is right for you.
- A good or even strong recommendation letter (the longer the internship the better).
- A possible publication.
But even finding a research internship is easier said than done! How can you approach this? My next blog post will deal in detail with the topic of how you can improve your application file for the application cycle in the next year.
Realistic School Selection
You should apply for about 10-15 universities. If you apply for more, you run in the danger that you will not have enough time to really polish your applications. If you apply for less you run into the danger of not being accepted anywhere.
You should have one or two backup universities where it is likely that you are accepted (> 75%). Often the university where you already studied at is a good candidate for this since your recommendation letter writers will be known to the university faculty. Apply for all top universities where you have some hope of getting admitted (>10% chance). Fill out the rest of the university slots with universities where you expect to have a good admission rate (25-33%) — you should have a minimum of 3 universities of this kind. These universities are usually the ones where a recommendation letter writer has a personal connection to a faculty with whom you would like to work.
Note that the best advisors are not necessarily at the top schools. You can get excellent PhD training at many schools outside of the top 20. However, if you thinking about an academic career then the school rank will be really important and you should try to find an advisor at a top school.
Pick universities mainly according to possible advisors. Make sure each university has more than one advisor you would like to work with. Do not apply to a university where there is a single good advisor. If your list is too small, broaden your area of interest. For example, if you would like to do deep learning and NLP and you cannot find enough fitting advisors consider also some advisors in computer vision or other fields.
General Q&A
4 year UK PhD vs 6 year US PhD
In the first 1-2 years of a US PhD you will do quite a few classes since the US PhD is designed for bachelor students. On the contrary, the UK PhD is designed for students that have already a (1 year) master degree and will have few classes. Thus you can get started immediately with research in a UK PhD which can be a nice advantage.\
US PhD:
- Designed for bachelor students
- Classes for 1-2 years. Classes distract from research.
- Funding guaranteed with admission, that is, you have guaranteed positions as a research assistant or a teaching assistant.
UK PhD
- Designed for master students
- Classes for 0.25 – 0.5 years. You can focus on your research from start to finish.
- Funding can be problematic and is often dependent on your advisor. This is why it is important to get in touch with your potential advisor before you apply.
- Less prestigious (in most cases) and thus it will be more difficult to get academic positions after your PhD. It will be more difficult to get oral presentations, best paper awards etc due to visibility bias.
Also be aware of local effects. If you study in the US you will also be in a US research bubble. Same is true if you study in Europe or Asia. For example, researchers in Europe know the “famous” researchers worldwide, but beyond that, they know more European universities than US universities in general (e.g. Stony Brooks vs University of Sheffield). Same is true for other locations. If you want to join academia in Europe, and you cannot get admitted to top US schools, it might make sense to apply for mostly EU universities.
Is a master required for a PhD?
In continental Europe, bachelor degrees are usually 3 years long and you require a master degree to start a PhD. In the US and UK, bachelors are often 4 years long and you can start a PhD right after your bachelor.
Does work experience matter?
It can help especially if you work at a prestigious institution (Google, Facebook, McKinsey, Goldman Sachs etc.). Other work experience can help if it is software engineering related, but any research experience (research internship) will be seen as far superior. Just a good job and no research experience will not help you.
How to pick advisors?
- Look at recent publications to get a sense of overlapping interest. Avoid working with academics that did not publish papers recently. There does not need to be an overlap in current research, but you should be interested in the research that the advisor is doing.
- Look at the list of students that graduated and where they are now. If you cannot find a list of students that graduated this is a red flag (or a new faculty). This is a good indicator of the quality of advice and training that you will get.
- Does the advisor has a startup? How many students does the advisor have? The combination of these factors is a good indicator of how much time you can the advisor to have. Dependent on how experienced you are in research you will need an advisor that has more or less time.
- Is there a fallback option in the same department? Sometimes relationships do not work out. Protect yourself by having a second advisor option as a fallback.
Should one even do a PhD?
If you want to work in academia you will need a PhD.
In industry, everything is regulated by supply and demand. The supply of AI researchers will rise sharply in the next years. If the AI hype collapses the demand will recede. The situation might be very similar to the situation that data scientists face in 2018: Companies only take over-qualified applicants because there is much more supply than demand. In this situation, a PhD will make a big difference if you want to switch jobs or want to be promoted. You might get hired without a PhD now, but without a PhD but you might have problems if you want to switch to another research lab (because the supply of skilled PhDs might be high, while demand is low).
If the AI hype does not collapse (unlikely) then you can find and switch jobs easily without a PhD. However, note promotion might still be more difficult and you might need to do more “research engineering work” compared to research. If you are happy with a research engineer position a PhD might be useless for you.
Do not do a PhD for the reasons above alone. If you do not want to do research do not do a PhD.
Contact advisor before application?
This can make sense if one recommendation letter writer can introduce you to a potential advisor. However, this is not required in the US. It can also backfire since it removes a shroud of mystery around you and sometimes it is more impressive to see your publications and recommendation letters first rather than to talk to you in person and seeing the recommendation letters afterward. In the EU it is sometimes required to contact a potential advisor before an application. If you need to do so, also try to get introduced via someone that knows your advisor personally, for example, your bachelor or master thesis advisor. If you do not have a personal connection to your personal advisor you might want to write an email with:
- Your current advisor
- A sentence about your past work (optionally: where did you publish your work?)
- 4 bullet points about potential work that you could do with the advisor in the form of “idea: One sentence that explains the idea”
It is very unlikely that your potential advisor will read and even reply you if you do not have a personal contact. If you do not have a personal contact and you apply to EU (UK) universities, then you might want to apply somewhere else.
How to pick a topic for your research proposal?
The topic for the research proposal does not matter. Nobody will ask you to do the work that you described in your research proposal. You can pick your research proposal topic based on how easy it would be to reuse it across different applications. If you do not need to rewrite it for different applications you save a lot of time. One thing to consider: The more familiar you are with a topic the easier it is to write a good proposal.
Tangerine says
Hey Tim,
Do the universities in NA (USA, Canada) interview all candidates they hope to admit for a PhD ? Or is it common to get admitted for a PhD without an interview in these universities ? Particularly UofToronto, CMU, UT Austin , TTIC .
Tim Dettmers says
Only about 1/4 of all universities that admitted me did interviews before they admitted me. Hearing some experiences from other students seems to indicate that interviews are more common now compared to when I applied though.
Armin Bazarjani says
Thanks for your response Tim. You are doing everybody a great service with your blog, so keep it up!
EJ says
Hi Tim,
I found this post very informative – thank you for writing this out all so clearly. I am having a really hard time gauging whether or not to apply to a ML PhD program.
I have an undergrad in economics (3.9 GPA, 4.0 in all math/stats classes through linear algebra) from a flagship state U with a similar ranking to UW. I started out my career in the non-profit sector abroad working as a data analyst and gradually began self-studying/applying DS methods at work. After working for ~3-5 years abroad, I was hired as a DS at a start-up that primarily implements projects abroad.
At my current job, I have two professors who are supervisors that could write strong LORs (one from a non-top tier uni and one from a top 3 ML program). I am a co-author of one paper, but it was published in Frontiers. I also have another applied paper submitted but to a domain journal. Right now, I am working on two research projects that are applying more cutting edge/SOTA techniques that my recommenders will be able to speak to.
I primarily want a PhD because I am much more interested in research than data engineering and I am worried about not being able to move forward in my career without a PhD. Everyone else on my team has one. My dream would be leading an applied research team in the public sector. Am I competitive for a PhD program? Should I be looking at a masters first? Is there anything else I could do to increase my odds – like waiting to apply until a concrete paper comes out of these research projects or trying to enroll in additional for-credit classes in comp sci/math?
Thank you in advance!
Tim Dettmers says
Hi EJ,
I think you have excellent qualifications for a PhD program. While your profile is not strong enough to get into the very best programs, there are many professors that like students with industry experience. If you can get past the initial application committee, it is very likely that can find great professors that would want to take you on. This can even be at top universities and programs if you get past the initial hurdle. Your academics at work might be able to make your application more seen, so focus on applications where these contacts have connections/work at. So all in all, you can expect the application process to be a bit random, but you have good chances of getting into great PhD programs. I do not think a master’s is necessary.
M says
Hi Tim,
I had a question about recommendation letters. Specifically, I’ve been out of academia for some time, and I’m unsure who to ask for recommendation letters. I have industry experience, so I can ask for letters from research managers with PhDs and/or extensive publication records, but I don’t currently have a professor who could give me a recommendation. Would this be a significant issue?
Thanks for this article. It was immensely helpful for me.
M says
Oh, for context, I have a couple publications from my industry experience with maybe a couple more before I begin applications next year. I may have very indirect connections with professors (i.e. from online collaborations), but I’m not sure how to ask for a letter from people I’ve known purely through online collaborations.. Is that a common practice, even? Either way, the people I have the most contact with in the office have been industry PhDs.
Tim Dettmers says
With this additional context, I think actually your letters are fine. I am not entirely sure what online collaborations means, but if it is mostly slack/email + some zoom meetings that is more than enough for a letter. I would probably ask for a letter from one of these professors. Even if the letter is not as strong as your industry letters, it sends a certain signal that you worked with academics on projects. People just want to see some context/contact with academics to know that you are not too stuck in an “industry mindset”. Looks like you have a great profile overall, good luck!
M says
Hey! Thank you! Just wanted to say, your feedback was so helpful for me, and probably for lots of others who don’t have direct connection to academia anymore. I have no resources other than to ask people who are close to academia, and you’ve been so helpful to me/many in the same circumstance and hope you know that 🙂 Thank you.
Tim Dettmers says
Hi M,
A letter from an industry researcher is great. In my application, I had one academic letter, one industry letter, and one letter from an instructor (researcher I only had classes with). I think it is important though to balance your letters with at least one academic letter. It would improve your chances to get admitted by quite a bit. If you can have 2-3 letter from researchers in industry though, it might be that no academic letter is needed. I guess the mass of industry research letters would make up for the different environments. In any case, you can always apply and see what you get and improve your application for the next year if it does not work out. Research experience with researcher in academia might then be on top of your list of how to improve your application.
David says
Hi Tim,
This is a terrific post, thanks for sharing your experience. And also your comments are very insightful.
I know you are pretty busy and I really appreciate it if you could give me some advice.
I believe my profile is fairly unusual. I started my career in the army (outside the US) and then dedicated 10+ years of my life in building software-based startup companies. Although my ventures weren’t anything extraordinary, they gave me some financial freedom to a certain extent. I did lots of web programming and also led a team of 14 technical people in one of them.
I’m 35 now, and I decided to return to college. I’m going to the second year in Software Engineering in one of non-top Engineering schools in Canada. My GPA is not perfect (3.9/4.3), but I’m working on it.
I’m currently working for an AI Lab from one of the Professors this summer, and I found myself in research. I have been studying graph neural networks and recommender systems.
I aim to finish my undergraduate and apply to MSc/PhD at top schools in the US and Canada. I’m primarily focusing on learning and conduct my own research in order to publish. I’m wondering what I should focus on more in the next few years to have a chance to get admitted? What universities do you think my profile would fit better, if any?
Thank you so much!
Tim Dettmers says
Hi David,
I think your profile and path through life is great! I have seen that some people with a military background were discriminated against and I hope that your track record in tech makes people not look at that. You might just want to hide that part for the sake of the application. Otherwise, I think you are on a great path. One thing that I would very much recommend you is to work with other people instead of doing independent research. Independent research looks okay if it works, but more impressive is to have a recommendation letter from an experienced researcher say that you think have great potential to do research. So I would recommend you seek out experienced researchers to do research with. Since you seem to like your independence, you could try to find someone that is more hands-off. The fit of universities depends highly on your research profile. The experience that you have gathered is very valuable, but people want to see research publications and recommendation letters first and foremost. If you can have those, your additional experience will make you stick out and likely get accepted into good programs. So it is important to focus on publications and letters for now
RS says
Hi Tim,
Thank you so much for this insightful post! I was wondering whether my profile was suitable for a direct PhD admit or if I should do my master’s first.
I chose an affordable state university not known for its research, graduated with a 3.7 GPA with a minor in Data Science. I have had two ML-algorithm development internships at a FAANG and one SWE internship at an open-source foundation, Jupyter. I participated in a NSF-Funded Fellowship that culminated in a poster presentation at ACM SIGGRAPH, and a poster presentation at JupyterCon. Additionally, I did my senior thesis in applied deep learning with the applied physics group at NASA. I am currently submitting one-two papers at low/mid-tier conferences/journals, and maybe one at a top-tier conference. I can potentially get a letter of recommendation from an ACM-award winner and/or NASA-Ames research physicist.
Do you think it is worthwhile to apply directly to a PhD program for ML? Or should I go to a Master’s first to get some more publications? If it helps, I am in an underrepresented group.
Thanks again!
Tim Dettmers says
Yes, I think your application is more than sufficient to go for a PhD program. You will probably not get into the very top programs, but there should be many great programs that would accept you. A master can make sense if you really want to boost your application and get into the top PhD programs. But it is also a bit of a gamble, people expect a bit more from Master applicants compared to applicants with bachelor’s degrees. Because your application is relatively strong, it could make sense to go for PhD programs directly.
Naman says
Hi Tim,
If someone has bad undergraduation grade because of personal issues and taking graduate courses.
But have exceptional grades in masters from top institute in India. Along with research papers.
Does it help to overcome bad undergraduation ?
Tim Dettmers says
Yes, that helps a lot. Often, the undergrad GPA is used to pre-sort applications but for any application that makes it through this process, people only look at the latest degree + research experience. So for some universities, you might be automatically rejected, but there are many more than will see you as a strong candidate. You have the highest chances at universities that do not filter by undergrad GPA, which is usually the universities that are not overwhelmed with applications. So my recommendation would look for strong, underrated universities/programs.
Rama says
Hello Tim! I’m thrilled to have found your post. And also your comments, even though this blog post is already 3+ years old.
I am a senior year student in the Electrical Engineering undergraduate program from the top engineering school in Indonesia. My GPA is 3.72/4.00 (top 11.7%), I’m also the top 3 students in my major, and I have a TOEFL score of 103/120 (28,29,18,28). I haven’t taken any GRE yet, and I plan to fix my speaking score on the TOEFL test later this year.
I’ve just planned to go Ph.D. in the US at the beginning of this year because I want to pursue a promising engineering career in the US. I’ve just realized that there are no excellent hardware design careers in my country, so the only way to pursue it is to go abroad. I’m aiming to get Ph.D. in VLSI/ML/computer systems architecture at the top 20 US universities.
However, because I’ve just planned to go Ph.D. this year, I still don’t have any publications to write on my SoP or LoR. I recently signed up for a research program in Indonesia’s leading microelectronics research institute for about six months. My professor was one of our first local fabless company pioneers, but he seems to have limited connection with US Professors. I also have an intern with a Taiwanese Professor, but my condition it’s a bit “messy” with him. That’s why I think I can’t get strong LoR for me from both of them, at least to get into top US Ph.D. programs. However, there’s some hope to get another LoR from another professor in my university.
After doing some thorough research, I found a program to get an outstanding research experience with US professors. I could start at the beginning of the following year. If I do good in the future and impress my US-based advisors, I will have an international-level publication, and then I will have a strong LoR to apply. However, the problem is, it will take 1.5 to 2 years for me to reach that scene. Indeed, I don’t have any issues with it, as it could prepare my skill to do good research in the future.
So, there’s the question: Do you think I have a chance to get accepted in top US PhDs for the fall 2022 admission, or will it be better for me to do the research program with US professors first for 1.5 to 2 years? As we know, applying to US universities costs some money (~US$200), and it will be pretty expensive for my family and me to register in ten universities without a significant probability of acceptance. But, if I have the chance to get inside because my condition was met, why not? That’s what I’m confused about right now.
I really appreciate any help you can provide!
Tim Dettmers says
Hi Rama!
Thank you for your comment. You seem to be very driven it is just that people did not give you enough opportunities yet to prove yourself. I would try to apply for PhD programs and try to use the many diversity pre-application that are now available at many universities. At UW, we call this PAMS, and this process is already on-going, so I am not sure if you can still participate, but you could still write an email and see if you can get some mentorship for your application (which should also be useful for other universities).
In general, if you are comfortable with it, I would also recommend you try to go a longer path. I did this myself and it worked out well for me. It requires some patience but it can make the following years much easier. So if you can do the 1.5 – 2 years that would give you a great start along that path.
Some universities can waive the application fee for people that have difficulty paying for it. I would see which universities offer this. Otherwise, try to carefully weigh where you apply. I think for many top 20 universities your profile is currently a bit too weak. Since you are so eager, I could see it working out after spending another 1.5 – 2 years, but in the current state I would spend the application fees on some lower-ranked universities. There are some great universities that are underrated which you can more easily get into. Look for universities that have groups that do great research but where the university is not super well-known.
Good luck!
Armin says
Hey Tim, sorry to further inundate you with a “chance me”, but seeing how you’ve been giving great feedback to others, I thought I’d reach out too.
As a quick synopsis, I did my undergrad and masters at USC with my masters gpa being 3.8. I have experience with three separate labs here at USC, unfortunately none of them resulted in publications. However, I believe the profs will give pretty strong recommendation letters. And, my interests lie within computer vision.
Now, my situation is a little complicated as I’m in a relationship and have to stay in Southern California so the schools I’m applying to are UCLA, USC, UCR, UCI, UCSD, and Caltech.
Obviously the main hinderance here is my lack of publications, but do you think that it would be feasible to land a position at one of these schools or am I selling myself on a pipe dream?
Also fwiw I have a connection to UCLA (Professor Emeritus) who said he can put me in contact with some professors there.
Sorry this turned out a lot longer than I had intended, but I would love to get your input.
Thanks!
Tim Dettmers says
Hi Armin, I think you have a good chance of being accepted at these schools. Recommendation letters are much more important than publications. It is possible that you are filtered out early because you do not have a single publication, but for all remaining schools that have a closer look will see that you have extensive research experience. So I think you have a good shot to get into a least one of these universities!
Umer Qureshi says
Hi Tom, great article! I’m a Sophomore in high school right now, so a lot of this information doesn’t apply to me at the moment, but it’s great to keep in mind going into college. One question I have though is how does one measure the “rank” of colleges? I’m planning to go to Purdue University and while many say that it’s a great college, there isn’t a general consensus on its rank among other colleges. Thanks for all the valuable information!
Tim Dettmers says
I guess there is the “ivy league” and other colleagues deemed very reputable for example, MIT, Stanford, CMU, UW. Beyond that, one good indicator is to look at the research rank R1, R2, R3 which is a rough classification of universities in terms of how much research they are doing: https://en.wikipedia.org/wiki/List_of_research_universities_in_the_United_States
RA says
Hi Tim,
Thanks for a great post. It is very insightful.
I am planning to apply for US universities for Fall ’22 admissions.
I am in my junior year of Computer Science UG program in one of the old (top 5) IITs of India. I have a GPA of 9.51/10 and a GRE score of 337 (Q-169 and V-168). I’m yet to give TOEFL. I have done one CV internship with a ML startup last summer and will be doing one ML internship with Microsoft R&D ltd this summer. I am doing a semester long research course (project) in DL now, which may result in a publication (may be, only as a third author). I am planning to take one more research course (project) in the field of either DL / RL in the next semester too, but may not result in a paper before the application due dates. I am confident of getting strong LORs.
My ultimate goal is to do a PhD (CS) from a top university in US. However, I am not sure whether my application will be strong enough for a top CS PhD program in the area of DL / RL.
Can you please guide me whether I should apply for a direct PhD or take the route of doing M.S. first and then apply for PhD?
Tim Dettmers says
Hi RA,
I think the connections to top universities are important if you want to have a chance. I think you might have some chance if your recommenders have connections albeit it is still low. Admissions have been extremely competitive the past two years and I think the best path forward to get into a top university is to do a masters at a top university. A masters from other universities is also fine, but it is important to do research with people that have connections and you should demonstrate good research skills. Even then it will be very competitive so all you can do is to give it your best shot! Good luck!
Lisa says
Hi Tim,
Thanks for sharing. I was confident that I can get an PhD offer until I read your paper.
Currently, I study Data Science at Fordham University with Merit-Based Scholarship and Graduate Assistantship, and work with a Professor on Natural language processing and deep learning research projects (may have publication this coming summer). Before that, I graduated Magna Cum Laude with a double major degree in Finance and Mathematics in less than three years (GPA3.7/4 for finance and 4.0/4 for mathematics) at University of Alabama and followed that up with a MS in Actuarial Science GPA(3.7)at Columbia University.
I have worked as Actuarial consultant at New York Life and Numerix(fintech) company for 2 years.
And have at least 2 strong recommendations. But I haven’t gotten any publication yet.
I’m applying data science PhD at NYU, Columbia, Yale, CUNY, Stony Brook, and UCONN. I’d be much appreciated if you could give me some feedback and suggestions. Thanks Tim!
Tim Dettmers says
Hi Lisa,
Many professors love to work with some more mature students that already have work experience. However, for competitive programs, most students will have at least one paper. If you have strong recommendation letters that attest to your research skills, it could still work out, in particular, if people read your application in detail. The problem is, many universities now take having a publication as a filter, and people do not look at applicants that have zero publications (this is so, in particular for NYU, Columbia, and Yale). This means, if you have some connection to these universities (maybe Columbia?), it might also help you get in touch with people and let them know that you apply. That might help you not get filtered out, and if your application is read in detail, you might still have a good chance to get admitted.
Otherwise, you could try to do a research internship or a Master’s degree first and then do a PhD. The Master’s program at NYU is excellent, and many Master’s students at NYU are also admitted to the PhD program, which is otherwise very competitive. You can also try to get into a research residency program. Facebook AI Research in New York is an excellent program that would make it easy for you to get into PhD programs the next year.
Good luck!
Ankit says
Hello Tim,
I must appreciate your thorough work on guiding prospective PhD applicants.
I felt lucky when I first came across this website on how to process one’s applications.
I would really appreciate your help in reviewing my profile and providing your
expert suggestions.
papers/patents
1. Co-author in Nature Scientific Report ( did all the code implementations)
2. Co-author in IEEE TIE (assistance in experiments and code implementations)
3. 1st author in ICASSP. However, acceptance details will be announced on Jan 21, 2021 (after
application deadline).
4. I have 5 Patents.
Recommenders – expecting good LORs from all three.
1. My masters thesis advisor (Phd in Signal Processing from IISc – top research school in India).
2. My boss from the company I am currently working in (PhD in Physics from Purdue).
3. My super boss from the same company (PhD in Molecular Simulation from IISc). He is also the head of our research lab in Samsung R&D Bangalore.
Work Exp – Samsung R&D Bangalore.
1. I have 6+ years of work experience in driver development, system design, Kernel level programming, Android application, data analysis and model development, Reinforcement Learning, CV, Causal Analysis, Graphical Models.
2. I have also started mentoring two of my colleagues.
3. I would say I have 2+ years of research experience.
4. My masters thesis work was on causal inference in neural network. My masters programme (2.5 years) was sponsored by the company based on my past performance. So I can say I have some experience in handling classes, thesis work and office (research and development) work, as well.
5. I have also helped in ideations and enhancing research proposals.
6. I have shown effective team collaboration skills.
7. Awardee of research to commercialisation award and top contributor award.
School
1. Masters from Top 10 technical university in India.
2. Bachelors from a Tier-2 University. I secured 1st rank in CS department.
3. Won 1st prize in a national project exhibition among 300 projects during my bachelors.
Score
GRE : 303 (142/161/3)
TOEFL : 94 (23, 25, 21, 25)
GPA
Masters GPA : 3.73 / 4.0
Bachelors GPA : 9.29 / 10.0
Area of Interest
I am super excited to work in bringing causal analysis principles in machine learning domain and its applications. However, I am also fascinated to work in Computer Vision, NLP, RL and applied machine learning.
Questions
1. Do I have a chance of being accepted by top 20 universities in US for PhD programme ?
2. If 1 is yes, then which schools should I target/hope for ?
3. Any suggestions on structuring SOP or including any other details to boost my success ?
4. How much impact can my GRE/TOEFL scores will have ? Lots of universities have mentioned
to not consider GRE score for Fall’21 applications. Although you have said above that these are just a filter, still you have some words for my score it will really help me to see clearly.
5. Any other suggestions ?
Tim Dettmers says
Hi Ankit,
You have a wonderful profile. Many professors look keenly for mature students with a broad range of skills and experience that you have. Most professors would hire you on the spot! However, the admission committee is often more focused on publications relevant to their field. I think in your case, a personal recommendation by your recommenders could do wonders. What I mean by that is that if your recommenders know somebody at universities whom they can let know “that a great student is applying,” that could help you. If professors have a keen eye on a particular student, they can often affect the admission committee’s outcome (in some universities more than others).
Other than that, I would recommend you mention groups closest to the papers/patents that you published. Once you are admitted, you can look for advisors you are actually interested in. If your patents are also relevant to some research areas, it could also help write about them (along with your papers and thesis) in your statement of purpose.
1. Yes, but it might be determined mostly by personal connections and framing your interest in your application around your papers/patents.
2. The schools that your letter writers have connections to.
3. The patents might be a good addition to your papers. If you highlight the research process that went into these patents, this can look very great in SoPs.
4. Your GRE scores and TOEFL scores are okay. People only look at these if they are too low. The 142 might look a bit strange, and people might ask why it is so low in an interview, but I do not think any university would reject you because of that score.
I think overall, you have an exceptional profile. The problem is that some admission committees are very narrow-minded. If it does not work out this year, try again next year. I can see you being admitted to the best schools that there are, but also being rejected from all schools. In your case, it is likely very random. The important thing is not to lose hope in case things fail. Good luck!
Siddharth Srivastava says
Hey Tim, thanks for the insightful post!
I’d be grateful if I could get some feedback from you regarding my research experiences. I’m just finishing my undergrad from Maharashtra Institute of Technology, Pune, India. My GPA is ~9.9/10, GRE is 328 (167Q, 161V, 5AWA), and TOEFL 119/120. I’m applying for MS courses and I’m targeting the top schools (CalTech, MIT, CMU, Stanford, Berkeley, Georgia Tech, etc.) in the US, and a few safe ones in UK and Canada.
My research experience has revolved around physics and machine learning (and often the intersection of both). Although, I’ve been researching for quite a while, I haven’t been publishing papers in ML. I do have a white paper in Physics that won the best paper award at a national conference. Beyond that, some of my work has been in applied ML and has been deployed (both in terms of software engineering, and ML for social good applications). I had bagged a summer research internship this year at a premier research institute in Europe, but due to the pandemic, I couldn’t go, and that gig had to be deferred! Substituting that, I have been working on time-series, satellite imagery, etc. in applied ML with an advisor locally.
As you can imagine, considering my scores (and assuming SoP is okay), I’m just worried about the lack of publications (even though I have been researching actively, and am currently submitting 2-3 papers, subject to some final experiments). How big a problem is the lack of publications? My recommenders can attest to my research capabilities (although they’re not very well known globally since I come from a tier-2 college), and in general, I think I’ll get strong LoRs. My only concern is the lack of publications then. Do you have any advice for me?
Thanks a ton!
Tim Dettmers says
Hi Siddharth,
I think you are in a very strong position for Master’s applications. I would also apply to the CMU PhD program since they often try to find candidates that other schools overlook. I would prepare a CV which also highlights research and some of your applied ML work. Also, consider the NYU data science masters which is very interdisciplinary and might fit your physics and ML niche very well. I think the lack of publications are not a big deal since you have pretty good LoRs. LoRs are always more important than publications.
Even if it does not work out, be patient. It seems you are very dedicated and it is just a matter of time once you can get into very good PhD programs.
Good luck!
Siddharth Srivastava says
Thanks a ton, Tim! Really grateful for your feedback. Take care!
Andrew says
Hello Tim, thank you so much for the detailed analysis and post. I doubt that there is a better post out there on the web that goes into as much depth abt the admissions process for CS grad school as yours.
I did my undergrad at KAIST, the top science and engineering university in Korea from Mechanical engineering (ME), although I have taken 3 CS courses. My undergrad GPA is 3.54 overall, but in the final 5 semesters my GPA was 3.75, during which I was awarded the Dean’s List twice.
My ME background is very control-theory focused and I excelled in graduate-level control theory courses as well. The link between control theory and reinforcement learning is why I wanted to enter machine learning in the first place. These days my interests are not limited to reinforcement learning but I am also interested in finding ways to integrate classical tools from engineering and mathematics e.g. probabilistic graphical models, control theory, signals processing (e.g. Fourier transforms) into modern deep neural networks to make them more memory-efficient on certain tasks and/or provide some form of guarantees on either performance or stability. My dream lab is Zico Kolter’s LOCUS lab at CMU MLD / RI.
During the final 4 semesters, I also did research on the design of hybrid electric vehicle powertrains (related to optimal control theory and systems modeling – traditional ME stuff), which became the basis for my thesis and later a publication at a top-tier journal. I am now currently a research officer working on robotics (computer vision and path planning) at a military research institute in Korea, where I am serving as part of my mandatory military service (3 years).
GRE: V: 158, Q: 170, A: 4.5
TOEFL: R: 30, L: 30, S: 27, W: 27
Papers:
1. 1st author (accepted) at IEEE TITS (Transactions on Intelligent Transportation Systems), which is the top-tier journal in the field of transportation systems
2. 1st author (in review) at AAAI, the acceptance for which is announced on Dec. 1st (before application deadline)
Recommenders:
1. My thesis advisor from KAIST, who has known me for the past 4 years, was the corresponding author on my accepted journal paper, and did his PhD from U. Mich. Ann Arbor in mechanical engineering (Top 4).
2. The director from my research institute (my current employer), who has known me for the past 2 years, advised me in the initial stages of research that lead to my AAAI paper (sadly not an author for the paper), and did his PhD from CMU RI (Top 1)
3. Professor from Postech in EE, another top 2 science and engineering university from Korea, who’s currently advising me on another piece of research (not sure whether we’ll get sufficient results to submit to arXiv before the application deadline), did his postdoc at MIT (Top 2) and worked at Facebook AI as a research scientist (has connections at MIT and CMU RI).
Questions:
1. Assuming that my AAAI paper gets accepted, what are my chances for PhD at CMU RI / CMU MLD, MIT, UCB, U Penn ~ Top 20 schools? Furthermore, what are my chances for Masters at CMU RI / CMU MLD, Stanford, U Penn MSE in Robotics, Princeton, Colombia ~ Top 20 schools?
2. How would I weave any publications that are not strictly related to machine learning, such as my IEEE TITS paper in hybrid electric vehicle optimization? I know having any prestigious journal shows research potential, but I was wondering if there was a better way to leverage this research achievement in my application.
3. Similarly, how would I justify my mechanical engineering background when the bulk of the schools I am applying to are in CS?
4. How would the SoP for applying to Master’s program differ from an SoP for PhD program?
Tim Dettmers says
Hello Andrew,
I think you will have excellent chances to get admitted to many top universities. Making the dean’s list twice having strong publications and excellent recommendation letters with research experience at multiple institutions give off a very balanced and mature impression with strong research potential. I think many advisors would like to work for you because of all of these reasons.
1. I think with your AAAI paper, you should have a +75% chance at CMU, 60-80% for most other top 20 schools.
2. I think if you weave your experience there into a robotics angle could work out quite nicely. Everybody in robotics and other fields knows how difficult it is to do research with robots and other machines. Experience with manipulating such devices for experiments will look very strong to anyone working with real robots (and not simulated robots).
3. I think the best way is really to combine it somehow with robotics. Another angle could be a more mathematical approach that comes from a control theory perspective. Combining the math of mechanical engineering together with machine learning and reinforcement learning to solve control problems in robotics (and otherwise) also sounds very appealing. I think this could be very interesting for many potential advisors.
4. Not much; I would use the same SoP. I guess you will apply for master programs at research institutions anyway, and these often also offer a PhD after the master’s program — using the same should do fine and will save you some time.
Andrew says
Thank you for such a detailed response! I admit you gave me extra motivation when you told me I have a good chance of acceptance at the top 20 universities 🙂 I just wanted to follow up on some points you mentioned:
1. The reason I was hesitant on recycling my PhD SoPs for my MS applications was cuz the lingo in the PhD SoPs may be too technical for committee members overseeing MS admissions. Should I need to write my MS SoPs in a more generic manner, or it is ok to expect the MS committee members will understand a good portion of what’s going on in my PhD SoPs?
2. how would you recommend I spin my research experience that did not result in any publications? This is both problematic for my SoPs and my LoRs, as all of my recommenders requested a so-called “brag-sheet” that could help them devise their LoRs.
a) For instance, during my stint at the military research institute, asides from the AAAI research, my main job was participating in an international project (between Korea and the US) aimed to build a robotic platform for exploring unmanned environments. I was able to gather some valuable RL experience and experience coding a real robotic platform (mainly in C++ within the ROS framework), and we were able to demo an (incomplete) version of our algorithm to our international collaborators. However, this sadly DID NOT lead to any publications.
b) Furthermore, my current research being advised by the Postech professor (different school to my alma mater) has recently met a stumbling block, so I am not sure whether I’ll be able to upload even a preliminary version of the research paper to arXiv before the application deadline.
Tim Dettmers says
1. If you have time to spare, I would write an extra SoP for the Master’s. If you are short on time (too many applications etc.) I would probably use the PhD one.
2. Write in your SoP about what you learned. That your writers ask for a “brag-sheet” is normal and not an indication that they will not write in-depth letters. If you worked with them closely, that should be sufficient for them to write in detail about how strong they see you in research potential (research potential is more important than publications).
2a. This sounds absolutely great! Make sure your recommender knows about all these details and writes about it. Also, mention what you learned in your SoP. A publication is not necessary!
2b. Papers do not matter in this case. It is probably more important how you deal with the stumbling block. If you deal well with it, that is something that your recommender can write about and is a valuable research experience.
All in all, letters are more important than papers! So do not worry about the details. Write about your research experience in your SoP, and make sure your recommenders have all the information they need to write a great letter!
Good luck!
Andrew says
so it turns out that my AAAI paper did get accepted! but since it hasn’t really been published yet, I was wondering what would be the best way for the admission committee to know that I’m telling the truth?
– should I create a personal web page where I can host a preprint copy of the paper?
– should my recommenders vouch for me that I got my paper accepted?
– does the admission committee even care about the actual contents of the paper? (will they read it if I provide a URL link in my online application?)
Thank you
cranxter says
Hello Tim, Thanks for the detailed analysis and post.
My undergrad GPA is 67 (first class with distinction; top 15% in my class). I also did a course based MS at USC(GPA : 3.4) where I did as research for a year as a RA . I was unable to attend a final exam in one of my classes hence the drop, I have a legit reason for missing the exam.
GRE : 320
Papers: 1 ,2nd author at ISBI , 1 paper at ICCV and 1 paper in AAAI both first author. My recommenders are a Professor from HKUST , research scientist at Amazon(USC PhD) and another research scientist (PhD from NUS and postdoc at U of Tokyo). They don’t have personal connections but they publish a lot and serve as AC’s.
What are my chances at CMU RI , UWash , Berkeley , MILA , UCL (London) ,Stanford, Caltech?
Tim Dettmers says
Hi Cranxter! You have a strong profile overall. The GPA can be problematic. Highlight your top 15% GPA in your undergrad on your CV and explain the missing of the exam at the end of your SoP. Otherwise, I think your chances are around ~50% for CMU, UW, MILA, UCL, and Caltech. For Stanford and Berkeley probably around 25% (might be a bit lower since these schools weigh GPA quite highly).
cranxter says
Hello Tim,
thanks for the quick reply. I have read read answers on Quora By Prof. Ben Zhao and Prof. Jeff Erickson that explaining a low grade in the SOP is seen as a somewhat negative point. They go onto explain that the only way to overcome this is to retake the class or take another very similar class and do well in that to explain that you know the material well. I did that (the class I missed the exam for was a ML class ,I took a DL class and another different ML class and got an A in both). So do you still think I should mention the reason for missing the exam in my SOP?
Also I submitted another paper to CVPR this year as 1st author with one of my recommenders, do you think that will make a difference in my profile?
Mahtab says
Hi Tim,
Thanks for writing the article it has been
My Profile:
School : Bachelors and masters from on a top 10 university in India.(Top 3 in robotics )
Publications : 4
1 Joint First author In IROS
2 publications in IEEE IV one first author other second.
1 workshop paper in ECCV
GPA : is around 3.3
GRE : 321(154/167/3.5) TOEFL: 105(28,30,25,22)
Current: working In industry , work is mostly involves a lot of applied research similar to my masters work, might be able to get a publication.
LOR : expecting a strong LOR form my masters professor, another one form a prof with whom I have a paper and third one from my manager( he is a Phd graduate from CMU).
MISC: ACM-ICPC regionals
FUTURE: would want to work as Industry research after this, especially in the domain of Intelligent Vehicles as most of work and research has been around that.
Questions:
1) Do I have a chance of getting into a PHD in US, if yes then which colleges should i realistically target.
2) Was wondering whether targeting a MS now and trying to convert to a PHD later be better option.
3) How should I structure my SOP
4) You mentioned that applying without personal connections is futile for a european phd ? Should I even bother applying there as I don’t have any direct contacts there?
Tim Dettmers says
Hi Mahtab! Your profile looks quite good. Your GPA could be a problem but your publications outweigh the GPA by far. The problem could be that you are rejected purely on GPA basis and people do not look at your recommendation letters and papers/CV. If you make it past this automatic check you are having pretty good chances to be accepted to many universities in the US.
To lessen the impact of your GPA you could try a couple of strategies. If your last year of your master had a better GPA you could highlight this in your CV. If your bachelor/master was particularly difficult and you ranked well also list the percentile/rank (if available) on your CV. If you have a good bachelor GPA (and poor master GPA) you can also try to highlight that GPA (or vice versa).
With your profile, you could get into some Top 10 US universities if your advisors have a connection to them. Overall, I think it would be best to target Top 10 to Top 20 universities for which you should have 40-60% of being admitted. I would mix in a couple of Top 10 (CMU is often unbiased and will give you a fair chance) and some lower-ranked universities as safe-options. I would apply to about 10 universities.
Your SoP should be regular. If you have a good reason why your GPA is low you could list it at the end of the SoP. Otherwise, I would also target research labs that have some Indian researchers which indicate that there is less racism. But your profile is pretty solid so you should also have a good chance at universities and research groups that are biased against Indians.
Mahtab says
Thanks for your reply and suggestion, Tim.
I have a few more questions.
1) Recently I have been suggested that I should try for programs similar to MSR at CMU. Would you too suggest the same, apply to research oriented masters programes in Top 10 to increase my chances of getting in and later try to convert to a phd.
2) When We say top 10 colleges are do we mean top 10 in a general holistic way or top 10 in robotics specifically.
As for the GPA, I might be able to get two separate transcripts one for B tech and One for my masters. I had an Integrated degree so I had courses from both Electrical (btech) and Robotics (Masters) at the same time. The masters(robotics) grade should be around 3.6 or 3,7
Will definitely look at research labs.
Tim Dettmers says
Hi Mahtab,
1) You will have a great chance to get into any masters program, I think, but I think you should aim for PhD programs. Having a 2nd Masters is sometimes seen as a negative signal. On the other hand, a Masters at a university in the US gives you a lot of connections, and it common to see, for example, Stanford Masters students being admitted for a PhD at Stanford. I think it could make sense to add in a couple of Masters’s applications to “prestigious” programs. This will give you an edge for PhD applications after your masters.
2) Usually, it is more difficult to get into the Top 10 ranked universities than the top 10 ranked research groups for robotics. This is because admissions are often decided at the department level rather than at the research group level. Research group (advisors) have many say in admissions, but they are only superficially involved. If you can find an excellent robotics lab at a “less prestigious” university, this is also a good target since it is easier to get in, and you will have an excellent PhD education.
Joseph says
This is an excellent guide, Tim you did a good job with this. How difficult is it to get into a PhD program for systems? Generally speaking, how many publications do competitive systems applicants have?
Tim Dettmers says
Hi Joseph, I am not quite sure about systems. The requirements should be much lower since there is less competition. The best way to figure out what is needed is to look at first-year PhD students of system labs at universities that you are interested in. That will give you a pretty good guide on what is needed to get admitted.
anzo says
Hey Tim,
Nice e-meeting you and many thanks for the insightful post! Here’s my profile:
– School: UWaterloo
– GPA: >= 95/100, close to 4.0 (converting from 100 to 4 is a mess tbh…)
– GRE/TOEFL: decent but not top notch
– Internship experience: a mix of research (3) and industry (3) thanks to the co-op program
– Publication: 1 in cryptography — International Conference on Information Security Practice and Experience, to be exact (3rd author though :/)
– Projected LoR: 2 research supervisors agreed to write for me; the third one is an industry research and the prof is just too busy to reply my email request so I went for another prof that I simply took a class with.
– Misc: Putnam 3-time HMs (top 100), IMO 3-time medallist (though idk if people care since IMO is >= 6 years ago), ACM-ICPC regionals
– Targeted field: the “mathy” part of ML (optimization, statistical aspects) while keeping open eyes to the applications (vision, NLP, robotics, health). Will go primarily for PhD given that MS doesn’t really provide funding
I’m wondering:
1. What type of schools should I even apply? Tbh how I see applying to top 4 CS schools is the same as buying a rather expensive lottery that I don’t even put real hope on. My plan now is to apply to a range of schools (including Waterloo), might actually end up with 15 of them (poor recommenders have to fill out many forms 🙁 )
2. How would publication in a non-ML field be viewed upon? (I wouldn’t expect it to have the same weight as a NeurIPS paper but idk if that will put me the same into the “no publication” pool)
3. What are the types of schools that would value the achievement of math/programming competitions?
4. Just curiosity: when people say they have strong LoR how do they even know it? (Given that you aren’t supposed to read your LoR)
Thanks for even reading!
Tim Dettmers says
Hi Anzo,
It looks overall, you have a pretty strong profile. In your case, I would not worry too much about publications since you are pretty well-rounded in other areas. I would recommend applying to schools though that have research groups with are a bit more mathy. Otherwise, you can always apply for something closer to your publication, and once you are admitted, you search for ML advisors.
As a 3-time IMO medallist, some schools will be very interested in your while others do not care. For example, I know that Stanford cares about IMO medalists and sees it as a strong signal for potential research ability. It is difficult or impossible to figure out which schools prefer medallists, so I would not think too much about it and put your time into applying more broadly.
1. Do no worry about your recommenders. 15 schools is a good amount in your case. I would apply for a mix of top schools (7x top 10), medium schools (5x top 20-50), and some safety options (2x).
2. People will care more about recommendation letters than your publication if it comes from a different field. They take it as a signal “ok, he published something,” but not more than that. That is why getting admitted through cryptography and then find an ML advisor during visiting days can be a good strategy.
3. See above.
4. Usually, you know how well you did when you worked on projects with professors. You also have an impression of how well you did in class or what attitude your letter writer has for you.
Anzo says
Hey Tim,
Really appreciate your thoughtful reply!
Just a follow up on your advice on “apply to crypto and find ML advisor”, how would that work since these two are very different fields? Their similarity will be the mathematical aspects and theoretical CS (like asymptotic runtime) so I guess you might mean something like “apply to theoretical CS (or similar) and find ML advisor” (which is possible in some universities since some mathy ML faculty members have affiliation with other departments like stats or even math departments)
Chunshu says
Hi Tim,
I find your article to be super helpful before my application for a CS PhD program. I’m currently a third-year (final year) master student in a university in China that rank around 10th nationally and I’m applying to a PhD in the domain of NLP.
I think I’ve got good research background (>5 first or co-first authored paper in top-tier conferences such as NeurIPS, ICLR, ACL, EMNLP, and AAAI) and my letters will be good enough (from MILA, USC, and Microsoft Research Asia, all referrers co-authored paper with me).
My concern is that: First, My GPA is not very good (3.47/4), that’s because the grading in my university is kind of strict, my ranking is around top 15%-20%. Also my major GPA is better and my senior year GPA is good (3.75/4, top2%) whereas my GPA in first year is much lower. Second is that the current visa status in China is not very good and I’m not sure if it will influence the admission committee.
I’m planning to apply top 10 schools in the US and also some top schools in the Europe such as ETHz and EPFL. It would be great if you can comment on my concerns and some estimation about the chance I could get in to top 4 + UW! Thanks in advance!
Chunshu says
By the way I am also curious about what would you thinking about whether to pick US schools versus Europe schools for doing a PhD on ML/NLP? Thanks a lot!
Tim Dettmers says
Hi Chunshu, you are right that your application is overall very strong compared to your GPA. There is a point where the GPA becomes meaningless because the applicant is so strong in research. I think this applies in your case. Also, read this stackexchange post which is very relevant.
In general, I would highlight in your CV your GPA in your senior year and also always include the percentiles. There are some schools that are known to be very difficult and admission committees usually know what a good GPA is at a certain university. If you include percentiles this also makes it clear in the case your university is less known. It could also be helpful to mention in your statement of purpose why your GPA is low (for example because you spend more time on research projects) but only do this if you have legitimate reasons which were unavoidable and gave you a disadvantage. If your classes are more difficult than other universities, you can just mention it in one sentence (or just leave it out).
I think if your application is not sorted out due to automatic GPA cutoffs (a common cutoff is 3.5) you are very likely to be accepted at any university that you apply to (for Top 4 + UW it is probably +80%). I do not think that admission offices will reject you based on your chances of getting a visa (I think it falls under illegal discrimination in the US and is not allowed). In the case that Trump gets reelected, you might still want to consider some universities outside of the US if you feel uncomfortable with the xenophobia (with Biden, it would probably get a little bit better). I think even if you do not go to those universities, it is also useful to get a perspective of how things are in Europe and Canada (research in Europe is quite different). It is also helpful to make some connections to those universities that you can use later.
Overall, it can be a bit random which university rejects you due to your GPA, but any university that does so certainly changes themselves short — I think any university than can recruit you can see themselves as lucky!
Good luck!
Chunshu says
Hi Tim,
Thanks a lot for your advices!! They’re super helpful for me to prepare my application!! Thanks and good luck for you!
Shiva says
Hello!
This was a really useful read and thank you for spending the time to write it.
I will be applying for thesis based Masters’ program in Computer Science in the US and Canada for Fall 2021.
My background:
GPA:
Major in Instrumentation and Control Engineering, Minor in Computer Science at a top ten university in India with a GPA of 9.25/10.0 (3.7).
GRE: 168Q 160V. Haven’t taken the TOEFL yet, but I think I will get around 110.
My interests are in DL for Computer Vision as well as RL since it aligns pretty well with my undergrad major. I do not have any publications yet, but have submitted a paper for review to an IEEE journal, but I do not think I will get a decision before Dec. This work was on designing a network for a particular vision application and I am the first author. I am working on another paper which I shall be the first author for a RL application in process control and might have it submitted before the deadline to an international conference.
I have done research work with all the professors that I shall be getting the 3 LORs from for about 4 months or longer, so I guess I shall get good/strong letters from them. However I do not think that the professors are very popular in the ML community or well known internationally.
I have a mix of ECE and CS programs in my list right now since I assumed it would play to my background that I get from my undergrad major. I am currently in a dilemma about whether I am being too ambitious or if I should aim a bit higher. They are CS at U of British Columbia, CMU, UT Austin, Purdue, McGill, Texas A&M and Georgia Tech; ECE at CMU, UIUC, ULCA and EE at Stanford. I am considering including USC CS to this list. What do you think my chances would be of getting an admit among these given my non-CS background?
Thanks a lot!
Tim Dettmers says
Hi Shiva, what I have seen before that quite a few universities have a bias against Indian applicants in particular if they do not have publications. The students that can get around this bias often have some connection to these universities and multiple published publications when they apply. If you have a solid connection through your advisors then you do not necessarily need publications. So your advisors might not be well known, but they probably have some specific faculty that knows them at other universities in the US. Your best bet is to apply for those particular universities. There are also universities that are less racist and give anyone an equal chance like CMU — you will also have better chances of applying to those universities.
In general, your application is probably pretty strong for many universities that you list with a chance of being admitted around 30-50% (except Stanford which is lower). I would add a couple more universities at the lower end of your list as a safety option. Otherwise, you could also try to go for master’s degrees or internships which will make your PhD application very strong (+60% for each university). Good luck!
Jerry says
Hi Tim,
Thanks for the detailed guide to applying for PhDs, I found it a very engaging and interesting read.
I’d hate to take up too much of your time and ask you to do something so repetitive, but seeing that you are currently studying at UW, which is coincidentally my school of interest, I’d really appreciate if you could give a general idea of what I might need to improve in my profile in order to have a good chance of being accepted when I apply next year.
I’m a junior at the University of Waterloo (coincidentally also called UW in Canada) studying Software Engineering (basically EECS in the Canadian university system) with a double major in Math (specialization in Statistics). My GPA is probably going to be around 3.9/4.0 based on the GPA conversion tool provided on the PhD admissions website and my GRE scores are good (164V/168Q, 5.5AWA). I have one publication at the moment, which is in a journal on discrete mathematics with the topic being graph theory, and I have a second-authored paper that is projected to appear in a top ML/NLP conference next year, but may not be published before applications are due.
My worries mainly concern my publications, as your breakdown of the factors affecting admission chances appears to put a bit more weight on publications, which I feel I am missing to a degree. I also have research experience, however they are primarily with the same professors at my school, which I feel might put me at risk when it comes to finding a third LOR. I have a good amount of industry related experience as a software engineer, with experience at a large tech company, but I don’t find that getting a LOR from anyone in industry will be helpful. My issue is that I felt burnout working in industry and decided to pivot to research a little late, so I’m concerned that makes my profile stand out in a negative fashion compared to those who have been pursuing research all along.
How do you feel I should improve my profile, if you have any advice to give? I very much wish to be able to attend grad school, but I feel a little hopeless sometimes when I see the amount of research and publications others have. I know I can let my grades falter a bit in order to do so, but time feels very short and I’m not sure how to approach my situation.
Thanks!
Tim Dettmers says
Hi Jerry,
I think your profile is pretty solid! A lot of advisors are seeking strong engineering + math backgrounds, which is not that common. Some professors really value the maturity and experience of people that worked in industry before. I myself started quite late in research (for similar reasons), and for me, the industry experience also always was an advantage. I feel like the story of being dissatisfied in industry and finding that research is your passion would also work very well in your statement of purpose.
One publication is often enough at UW, especially if another publication is on the way, and your advisor can write a good letter on your newest paper.
I personally had one research letter, one industry letter, and one teaching letter, and it worked out very well for me. I would try to get an industry letter that attests to your engineering skills.
It can feel overwhelming to see other students have so many publications, but in the end, that does not matter so much. Publications seem so important because it is the “fuel” for recommendation letters. In reality, recommendation letters are the most important piece of evidence for your research potential. Often, it does not matter if you have 1 or 3 publications since your letter of recommendation will not be much better. Your advisor will already have a pretty complete view of you after 1-2 publications, and that is exactly where you are now. So, having additional publications would not really improve your application.
Overall, your application seems to be quite strong. I would think that your chance of getting admitted to UW is about 60-75%. I know a couple of professors who would fight to get an experienced, well-rounded student like you are — so no worries.
Jerry says
Thanks so much for the reassurance.
Just a quick followup on your comment about LORs. When it comes to getting an industry letter, what should I aim to have in terms of content of the letter? I would expect that speaking at length about work ethic may be redundant because that can be described in both a research and teaching letter as well, but speaking about problem-solving and learning skills also feels like something the other letters would contain. Do you have suggestions on what I should aim to have my past supervisors speak on, if I decide to get a letter from one of them?
Upita says
Tim,
Excellent article ! Gave me wonderful framework on how to approach things.
Many universities have waived of GRE requirements this year due to COVID. Do you think it will be easier to get into some schools or does that make it even harder ?
I am 35, and considering going back to get a Phd in ML/CV/NLP related topics. Do you think the age factor will go against me ? I suspect it will affect somewhat, but not sure how much that impact will be.
My undergrad is from an average university in India and Masters from ~50 ranking university here in the U.S. I did much better during my masters and got a 3.7 GPA, contributed as 3rd author in 1 paper publication at a Top conference(supercomputing) and 2nd author on a technical report.
I worked in the Industry(not top tech) for several years after that and have 3 patents(with others) in the field of autonomous driving. During the time I took few courses at Stanford to study ML topics. Does that help to show I continued learning even after I was working ?
I will get 1 good LOR from Professor with whom I worked with during my Masters, but other two will have to be from my Managers in the Industry. Is 1 academic + 2 industry LORs good enough ?
How good do you think are my chances of getting into a Top 10 university ?
Thanks,
Tim Dettmers says
Hi Upita,
I do not think waiving the GRE requirements will make it easier or harder. GRE was usually used to remove the lowest scoring applications automatically (without looking at them) to save time. People might now use different criteria for that such as GPA.
35 is a good age. Many professors enjoy working with more mature students that already have industry experience. It will not be a disadvantage in your applications.
Your research experience is spread a little thin, but might be just sufficient. I have seen that some academics view supercomputing as strange (because most universities have no supercomputing departments and they do not understand it), and technical reports are not valued in some fields. So it might be that people have difficulty evaluating your research potential. I think the most valuable pieces of your application are the patents. If you can get a recommendation letter highlighting these patterns and what research-like know-how and work went into them, it can look terrific! I would also highlight these patents in your CV. I would try to get one letter for your engineering skills, one letter focusing on those patents, and the research letter from your professor.
If you play it like this, you have a good chance to get into PhD programs in the range of Top 20 – Top 50. You can maximize your chances by mentioning research groups that work in either supercomputing or autonomous driving (or reinforcement learning, robotics, computer vision) in your statement of purpose. Once you are admitted, you can change your focus to ML/CV/NLP. I would probably get about 6 universities from the top 20 to top 50 range and 4 universities from the top 10 – top 20 range. It is best to apply to those top 10 universities where your managers and advisors have direct connections.
Good luck!
JS says
Hi Tim, thanks for this amazing guide which helped me gain a lot of perspective before applying for PhD programs. I am from a final year student at a top 10 university (NIT) in India with a Major in Production Engineering and Minors in CS. I am aiming for a MS by research and/or PhD program at top 4 university (CMU, MIT, UCB, Stanford). I have a slightly above-average major GPA of 8.3/10. My performance and grades in just CS courses are much better (9.5/10). I have had decent (but not top) research experience through internships. How much will having a non CS major and low grades (8.3/10) affect my chances for admissions (for MS by research and for PhD in ML)?
My research interests are in RL, statistical ML, Bayesian methods and optimisation
I currently do not have publications. The two most significant ones involved security for ML programs (the work was more on security but has applications in ML) and a RL application project for education (with a bit of novelty in the work, but not groundbreaking). The security paper is submitted (under review) with me as the 3rd author at an international conference. For the RL project, I am trying to at least get it submitted (if not accepted) to a conference by application deadlines (with me as first author).
I have 2 LOR’s in the good-strong area: a CMU prof and a prof in Univ. of Vermont (grad from PhD MIT EECS). Apart from this I have one LOR in the decent-good category. What are my chances for the MS by research and PhD ML programs in the top 4 unis given my Production engg background? How can I improve my chances if I don’t get into MS or PhD programs during applications this Fall?
Thanks in advance!
Tim Dettmers says
Hi Jithen,
I think you have a pretty good chances to get into a CMU masters or even CMU PhD program based on your LOR from a CMU prof. Otherwise, you got to play your connections. Any other institutions where your LOR writers have connections? Aim for those places. Right now you might be able to score some good masters programs, but I am not sure if top 4 beyond CMU would work.
Otherwise, I think most universities are happy if you have a list a little bit of CS on your transcript. It is also a bonus if your CS GPA is good. Highlight this in your CV and maybe even in the statement of purpose. Good GRE scores could also show that your 8.3 GPA was a “glitch”.
If your submissions get through before the application deadline this will help a lot, but it would still not be sufficient for Top 4 PhD programs. I would recommend to do a masters, get a strong background and then you should have a much easier time getting into Top 4 PhD programs. You can also try to get into residency programs which would also be perfect in your case!
Hope that helps! Good luck!
JS says
Thanks a lot for your helpful remarks! I decided to apply for 2-3 PhD programs (just to try out my luck) along with other MS programs I am applying to.
Update from my previous comment: I submitted 2 first-author paper (one at AAAI Student Abstract and one at ICLR). The AAAI SA results are out by application deadline, but the ICLR results are out only later. So at the end of applications I will have “submitted, under review at ICLR”. The AAAI Student Abstract paper may or may not be mentioned on my resume depending on its acceptance. I also recently gave my TOEFL and am expecting 110-113.
1. Does a ICLR paper that is under review have any impact on my decision or does it not matter until it is “accepted”?
Im planning to apply to the following schools for MS, some comments would be helpful on whether this is a good choice:
CMU, Stanford, MIT (MS+PhD), UC Berkeley, UIUC, UCLA, Princeton, Georgia Tech, McGill University with MILA, UMass Amherst, UWashington, NYU, USC Viterbi, Univ of Alberta.
2. Do you think this is a practical list for an MS or am I aiming for too many schools that are beyond my reach?
3. For Berkeley, what would be a wiser choice? To apply for MS or MS+PhD? I have no idea on how qualifications differ for these programs.
3. Do you have any safe/moderate/ambitious university suggestions?
Thanks in advance!
Tim Dettmers says
The paper under submission is good but will not have a major effect on your application. However, if your advisors mention the work in some detail that you did for that paper in a letter of recommendation, this is equivalent to having the paper accepted. What counts in the end is an indicator of academic potential, and an advisor can identify that better than conference reviewers.
1. I would mix in some universities with slightly lower tier also as a safety option. You will have a chance between 30% (Stanford, Berkeley) and 75% (CMU) for most universities. So if you go with those universities, you have a great chance of getting accepted into more than one university. Still, to make sure that you get accepted somewhere also applies, you have almost certain acceptance rates.
2. It depends. It costs a lot of effort to have that many applications, and there is a possibility that you are rejected by all of them, so a mix is better. Going to 7-10 great schools and 2 good schools would be pretty safe, and you might get lucky.
3. I would apply for an MS degree. It is unlikely that you will get into an ML-related PhD at Berkeley.
4. I think you have some moderate and ambitious universities already. You should add some universities where your recommenders have connections to which are less prestigious. Those schools often easily admit you.
JS says
I removed McGill/MILA and added UMich Ann Arbor which I suppose is still hard but relatively easier than McGill. Also, one of my recommenders have a postdoc @Umass Amherst which is already there in my list. Do you have any “safe” universities as suggestion? Do you think Purdue or UT Austin are good bets as safe unis? Based on my list, my thought was that NYU Masters in Data Science and USC Viterbi might be safe but correct me if I’m wrong (I haven’t applied before so my guesses could turn out to be awfully bad).
Tim Dettmers says
I think McGill/MILA is doable since you do masters before PhD and it is a bit less selective (as far as I understand). If one of your recommenders has a postdoc from UMass Amherst, that is also a safer option. In general, your diversification is already pretty good, and I would believe that it is pretty certain that you get at least one admission. Purdue and UT Austin go in the right direction, I think. You can add more similar unis for more safety.
The NYU masters is pretty selective, but I think it fits your profile quite well. I would give it a shot. If you get into the NYU masters, it will be much easier when you apply for a PhD. I am not sure about USC Viterbi, but overall your application is pretty strong for a master’s — so I would not worry too much about it.
JS says
Update: Got into the master’s thesis program at McGill/MILA which was one of my dream places!
Tim Dettmers says
Congrats! Happy that you found an excellent new home!
William says
Thank you for the guide!
Can I ask you about future plan because I want to get into top PhD program in CS (CMU, Stanford).
My profile:
– GPA 3.85
– GRE 168 (Quant), 165
– Research Publication (1 first author, expecting one more publication first author, before graduation)
– 2 strong recommendation letters 1 good
I want to apply for MS first because I am from Indonesia and there is no connection from my advisors. Where do you think I should apply first if I am aiming for top schools like CMU or Stanford for PhD?
Tim Dettmers says
Your application might actually be pretty good. Depending on the school, some would see an application of your strength coming from Indonesia as a strong indicator of great research potential. CMU is a school that might admit you right away to a PhD program. Other schools like Stanford are less forgiving as they care more about numbers. I think, however, that you could be well admitted to any top masters program. If you want to go to Stanford, the Masters program there is the best bet to do a PhD at Stanford. Otherwise, the NYU masters in data science is very strong.
In your case, I would also apply to other PhD programs. I think you can get admitted to many Top 20 programs although if you aim for better than PhD program than Top 20 you might want to do a masters first. After doing a masters, you should have no problem getting admitted to top schools.
Yaya says
Hi Tim, this guide is excellent. Thanks for publishing! Please have a look at my profile below and advise what can I do to improve my admission chances.
I intend to apply next year (2021) for PhD in ML at CMU to work with Ruslan Salakhutdinov and PhD in Social & Engineering Systems to work with Tamara Broderick at MIT.
I intend to do research in Nonparametric Models and Topological Data Analysis to extend Deep Learning with application to social phenomena.
I’m from Cameroon.
I graduated in 2014 with a Masters’ in Computer Science & Engineering (GPA of 3.28 /4) at National Advanced School of Engineering of Yaounde.
A little more about the above program: it is a 5-years program that is comprised of 2 years of courses in Mathematics, Physics and Introductory Computer Science followed by 3 years specialization in Computer Science & Engineering.
Coursework years 1-2: Real Analysis, Vector Analysis, Numerical Analysis, Abstract Algebra, Linear Algebra, Multilinear Algebra and Curves & Surfaces, Affine & Euclidean Geometry, Probability & Statistics, Mechanics, Electrostatics, Electromagnetics, Electrokinetics, Electric Circuits, Thermodynamics, Optics, Physics Labs, Introductory Computer Science, English & French Languages.
Coursework years 3-5: Theory of Computation, Compilation, Algorithmics, Information Systems, Databases & SQL, Programming, AI, Software Engineering, Distributed Systems, Boolean Algebra, Computer Architecture, Electronics, OS, Computer Networks, Computer Security, Project Management, Human-Computer Interaction, Measure Theory & Introduction to Functional Analysis, Probability & Statistics, Numerical Analysis, Data Analysis, Operational Research, Image Processing, Communication, Management, English, Internship, Capstone Project, Thesis.
I know it is too much (some subjects are split into 2/3 quarter course sequences), but all the courses are mandatory. 🙂
I have no paper.
I have little research experience working on Component-Based Software Engineering for my Thesis project.
I have MicroMasters in Statistics and Data Science from MITx completed in December 2019.
I have 6+ years of work experience in Software Engineering at a StartUp and Product Development at MTN.
I discussed with my Masters’ Thesis advisor who knows me very well (we extended our work in the Thesis scope to a consultancy project with the Government), he can provide a strong recommendation letter. But not sure I can get another one (at least a good one) from other faculty members. So I intend to ask for the 2 others from (senior) managers who really appreciate my work.
I started a long journey of 2 years (started this year) to review and learn prerequisites (I also love Math): Real Analysis, Linear Algebra, Point-Set Topology & Vector Analysis, Axiomatic Set Theory, Measure Theory, Abstract Algebra, Probability Theory & Stochastic Processes, Functional & Numerical Analysis, Mathematical Statistics, Algebraic Topology, Differential Topology & Geometry, Riemannian Geometry, Bayesian Statistics, Algorithmics, Artificial Intelligence, Convex Optimization, Theoretical Machine Learning, Statistical Learning, Probabilistic Machine Learning, Deep Learning and Topological Data Analysis. For each of the above, I follow a book and related course (videos, notes, exams, etc.) from the author when publicly available online. I already completed the first 4.
I intend to get practical knowledge of what I learned by taking Kaggle competitions from Jan 2021.
I intend to maximize in IELTS, GRE General, and GRE Math next year before applying.
My study plan ends in November 2021 so I’ll not have enough time to write a paper before applying.
Thanks
Tim Dettmers says
Hi Yaya, from your current profile it would be difficult to secure the PhD programs that you are interested in. The main problem is that the admission committee would have trouble to identify signals for research potential. The research experience that you got for working on your thesis would yield a good recommendation letter but the standards for admission now often require at least one published paper, especially if you are a masters student. Your GPA could also be interpreted as borderline. Some schools are very difficult and you could convince the admission committee that you did very difficult classes and your GPA is good if you could show that you were in the top 5-10% of graduates for example. Otherwise you would probably convince them through additional research experience. The GRE Math could also be useful to show that your GPA is not comparable to a US GPA. But on the other hand the GRE Math is hard and your time would be better spent doing more research if you have an opportunity to do so.
What I would recommend for you is to either get more research experience through an internship abroad or collaboration from a distance (which should be easier to do during COVID). If that does not work you can try to do more research with your master thesis advisor and do the GRE Math. In the second case, I would apply to schools between top 20-50. In the first case, it would depend on how much research you can publish. You might want to gather more research experience and then apply in in 2022. This way you might be able to get into some Top 10 schools.
Good luck!
Dipro Ray says
Hi Tim, thanks for the great post! I was wondering how important publications are for MS admissions?
Some background about myself: I’m a CS major and Math minor at UIUC with a 3.99/4.00 GPA (A+s in nearly all CS, and all Math courses). I have one year research experience at 3 different labs (areas: bioinformatics, data mining, scientific computation). By the time I apply, I will have completed 3 internships (2 Google/Facebook, 1 top unicorn), of which one was ML engineering related and one will probably be on an AI Research team.
I’m confident about getting good or strong recommendation letters from my advisors (and in case it matters, 2 of them are Stanford and Berkeley alumni respectively who publish in top conferences each year in their respective fields). One of my advisors is also supervising my senior thesis (which I’m completing my junior year). I also anticipate a pretty good score on the GRE. I hope to get into a top (Stanford/Berkeley/CMU) MS program.
Problem is: I do not have any publications yet. I am sure that my work will eventually result in them, but I’m not sure if it’ll be in time for my grad school applications (one of my projects is still ongoing; for the other, my professor wishes to attain even better results before publishing). What chances do would you think I have at an MS acceptance (with my senior thesis, and hopefully at least 1 paper submission)? In case I find myself unable to produce papers or paper submissions by the time I apply, do you have any suggestions as to what else I could be doing?
Tim Dettmers says
I think in your case publications are not that important since your recommendation letters will be strong and you are well connected. Definitely apply to Stanford and Berkeley PhD — you probably have about +30% of getting admitted. With a single publication that will probably rise to +60%. So really try to get something published before you apply. In terms of masters you will have it relatively easy to be admitted to Stanford, Berkeley, and the institutions closest to your advisors. For other masters programs it is less likely that you are admitted but it should still be +60% and if your advisors have connections it should be +85%. I would aim for top masters programs and otherwise PhD programs to which your advisors are connected. With your profile and a connection from your advisor your chance of getting admitted to Top-20 PhD programs is about +40% with no publications. Make sure to select some master programs in the Top 20 as a fail-safe. I would not apply to bad PhD programs since you are likely to move from your masters directly to a great school (just make sure to keep up your research and specialize in a certain research field).
adas says
Hey Tim, how difficult is it to get into the Allen AI2 PYI program specifically for the PRIOR (Vision) group? Does research experience at university labs suffice to get into this program or does one need a stellar profile with multiple publications to get in like with other AI residencies?
Sam says
How would you compare the acceptance rate of ML Phd compared to CV or NLP.
Is it easier to get a CV PhD admit compared to ML or vice-versa?
LH says
Hi Tim,
Great post. I was wondering if you could take a look at my profile and rate my chances for Ph.D. admissions later this year:
BS+MS from a (below top 20, within top 100) school
1 very strong recommendation + 2 >=good (all well-published researchers but not particularly well-connected)
accepted top-tier ML papers: 1 first author paper + 2 second author papers
3.6 GPA + decent GRE, TOEFL scores
I have a few more drafts that I’m submitting this year as first or second author. Overall, I think I have a decent publication record, but I also know that my school and the fact that I don’t have any connections to top schools are obstacles.
I’d really appreciate if you could rate my chances of getting into top-5 or top-10 schools, or any other strategies you would suggest for my situation.
Tim Dettmers says
Hi LH,
I think you have decent chances. The GPA is currently borderline. Would you be able to increase it a bit? What you can also do it try to improve it in the last semester and write “GPA of 3.x in the last quarter”. An upwards trend will make people believe that GPA is not an issue and that you are “smart” and not only hard working. You can also mention why your GPA is lower, for example you probably focused more on research than on classwork and that is good but the admission committee cannot know unless you tell them (and show them convincingly that research is the true reason why your GPA suffered).
Often a GPA can be rendered meaningless if you publish enough. Nobody will look at someone’s GPA if you publish a lot. You are definitely coming close to this area but with the lack of connections and the stiff competition will be hard. For Top-5 schools I would give you about 15-20% chance. For Top-10 schools +40%. If you raise your GPA a bit and publish another paper then for school to whom your advisors have connections is about +80% and for Top-5 25% (without a connection) and Top-10 about +50% (without a connection). If you really want to get into a top school you could try to do an academic internship at one of these institutions or a industrial residency; these are also competitive but it can work out if your advisors have some connections that can make it work.
LH says
Hi Tim,
Could you take another look at my profile? Two more of my submissions were accepted since my last comment, so I’m currently at
3 first(or co-first) author + 2 second author
top-tier ML papers. What would you rate my chances of getting into top-5 or top-10 schools? I have submitted my applications to Ph.D. programs starting next fall.
Also, do you think I should email potential advisors even though I have no connections? I have a short list of professors I’d like to contact, and there is one specific professor whose research very closely matches mine, but their website asks applicants not to email during the application process. Could it make sense to strategically ignore this public request if there’s a very close fit, or would I just be annoying them?
Tim Dettmers says
Hi LH, I believe your chances are excellent. I think you will have a good shot at most top universities although it is always a bit down to luck in the end.
I would probably not email the professor to respect their wishes. If you have an advisor who knows this professor personally it could be okay for them to reach out, but probably it is best to keep silent. Often it can be a boon to be unknown and have a strong application. In this way, you seem more mysterious and the sudden realization and surprise to see that there is this previously unknown great student out there can be very impressive. This surprise factor vanishes if you introduce yourself beforehand.
Sam says
Can you please recommend some ways to get more research experience after Master’s? I know of the AI residency program , what are some other ways I can get research experience?
Tim Dettmers says
The best way to get more experience is to ask your advisor to connect you with other labs where you could do an internship. Residency programs can be a good opportunity if you have significant hacking experience or community presence and a little bit of research experience is lacking, but otherwise, they are usually too competitive if you are in a position where you want to gather research experience.
Sam says
Hi Tim,
Great post, can you please share how many publications (along with their venues) did you have at the time of your application submission?
Tim Dettmers says
I had a single author ICLR and first author AAAI publication. But as I said, publications are not as important as recommendation letters and you should not focus on publications over recommendation letters. Publications are a great side-effect of working with researchers that help you to improve research skills.
RG says
Hi Tim,
This post was very informative and realistic about PhD admissions and acceptance rate.
That being said, I am in a bit of a unique situation. I am currently a junior completing my computer science undergrad at a small state college (CSU), with minors in data science, Chinese, and statistics. I plan to have 2+ papers submitted by the time I apply, related to deep learning with genomics and natural language processing. I’ve had strong internships, doing research at a top-4 tech company (FAANG) two summers in a row, and at a very prominent open-source foundation where I worked with the co-founder. I also co-founded the first Artificial Intelligence interest club on my campus, and we work on cool hacks. I also anticipate scoring high on the GRE.
My GPA will be around a 3.7, which is low for an unknown state school. However, since I have 3 minors and other relevant experience, do you foresee this being a big problem?
What schools do you think I should realistically apply to?
Tim Dettmers says
Hi RG,
You should not worry too much about publications and GPA. It would be nice if at least one paper is published before you apply but your recommendation letters will be much more valuable anyway. The research internships are great as well as your open-source and leadership experience. The GPA is in the good range so it should not matter.
I would recommend you to apply both to top schools and some schools below that. Your case is where certain experience can easily get you into Stanford/MIT but it would also be as easy to overlook you and be rejected by the majority of top universities, especially if 0 published (rather than submitted) papers are on your CV. It all depends if reviewers will have the patience to read your recommendation letters. Your statement will also be very important in your case.
Overall, you can expect high variance from your application: You will have a solid probability for almost any school (around 50%) which means you can get into a top school or not. Just apply to many schools and reserve 2-3 universities for safety slots and you should be good. I am sure you will get into a university which will be great for you.
Good luck!
Minh says
Hi Tim,
Thank you for your post! I find it extremely helpful. I wonder if you can take a quick look at my profile below and rate my chances of applying to a top 50 Ph.D. program in ML/DL/CS:
I am a junior at Worcester Polytechnic Institute (#64 in National Universities US News). I major in Computer Science and Math with a GPA of around 3.5/4. I am the first author of a paper related to Deep Learning, which was submitted to ICASSP and is under review; and the second author of a paper related to Data Mining, which is still under preparation and we haven’t picked the conference for submission yet. I have been involved in research since my sophomore year during both school year and summer (I plan to do so next year as well). I believe I will have 2 strong letters and 1 good letter. I haven’t taken the GRE yet but in my first mock test, I got 168 Quantitative (94th percentile) and 158 Verbal (80th percentile).
Do you think I would have a good chance at a top 50 Ph.D. program? Also, do you think it is a good idea to also take the GRE Math Subject Test?
Thank you!
Tim Dettmers says
You have a pretty solid profile which would give you good chances for most universities in the top 50 (+60% for a large bulk of them). If you can get another paper, or round up your profile with a research internship or industrial internship in one of the big tech companies you can improve your chances still. The GRE math subject test can be helpful if you think about applying to theoretical machine learning programs/groups. Otherwise, I would invest my time more into other things like securing a good internship for the summer, to increase the quality of a paper or to get started with another research project. Overall pretty solid and you should have a good range of universities to choose from.
Shahzad Qureshi says
Hey Tim,
After reading this post and comments, I believe that I have the weakest profile but I am still daring to comment and putting forward my case in this comment. My bachelor’s cgpa is 3.21 and my master’s (opted course work) cgpa is 3.6. I had to go with course work masters because I had financial issues at that time. I was doing a full-time job to support my family and study as part-time. Because I could not find a suitable supervisor at that time and thesis seemed a full-time job so I chose the course work option at that time. Now I really want to do my Ph.D. in machine learning recommender system. I have 10 years of experience as a software engineer and my age is 34 currently. My IELTS score is 7 overall with each module band 6 +. I have 1 publication (software engineering related) as a co-author in IJCSNS, I even don’t know if this journal or publication is worthy enough to be mentioned in CV or not. I can get 3 recommendation letters though. Now if I publish 2 papers in tier 2 journals of machine learning, will I be able to secure a Ph.D. scholarship from Europe countries or Australian universities? Any tips for me?
Tim Dettmers says
Hi Shahzad, yes please do not mention your IJCSNS in your CV. This is seen as a predatory publishing journal and would look negatively. Your GPA does not looks not too bad as often the latest course work is evaluated (Masters 3.6). The story of financial hardship is very important and should be mentioned in your SOP. This is seen as positive and shows that you can push through very tough periods of high stress — an important quality for a PhD. Your software engineering experience is great. Age does not matter. The recommendation letters would determine your chances of being admitted mostly. If your advisors have some connections to some universities this is your best shot at being admitted.
Funding in Europe and Australia is complicated. I would not say that papers secure your funding. Often it is much more arbitrary. I would just apply and once you are admitted try to figure out the funding situation. You might also want to consider countries were doing a PhD is considered a job and yields a stable good income. For example, Germany, Belgium, and Switzerland have this system. You can also find better funding in Norway, Denmark, Finland and to some degree in the Netherlands. You might also want to consider PhD positions which are partly funded by companies. Companies would be very interested in your software engineering experience and this could be a great match for you. Overall, your profile does not look too bad for the right universities. Really try to exploit the connections that your recommenders have and you might get into a program that offers a great experience.
Sam says
Hey Tim,
Thanks for the post, it was incredibly helpful.
I’m a CS undergrad at a university with a somewhat recognized CS program (BYU). I’m aiming to do a PhD in computer science/cognitive science, and am shooting for top programs (Berkeley, Stanford, UCSD, Caltech, MIT). I have a couple of years of research experience in 2 different labs, although neither has been strictly CS research. One involves computational modeling of molecular dynamics, including using ML to analyze MD simulations. Through this work, I’m a contributing author on 1 paper, and a 1st author on another (that has been submitted but likely won’t be published before I apply). The other lab is bioinformatics, and I’m a contributing author on a high impact paper published in Cell. Additionally, I’m currently leading a project that I initiated (and for which we received a fairly large grant from the national institute of justice to fund) that aims to use ML to predict samples from sexual assault kits that will be most likely to contain DNA profiles, in order to aid forensic scientists in selection. This should be a fairly impactful project. I think I will get fairly strong letters from the professors I work with in the 3 areas mentioned above, although as my research has been fairly interdisciplinary and outside the area I want to pursue in grad school, there probably won’t be connections between my recommenders and professors I’m interested in working with. My GPA is 3.99/4.0, with 162V/169Q/5.0 GRE.
What do you think my chances of being admitted to these programs are? What could I do during the application process (i.e. in my SOP) to improve my chances?
Thanks Tim! Really appreciate the information here and your feedback.
Tim Dettmers says
Hi Sam! I think your research background is fantastic. You rightly recognized that the lacking connections between your recommenders and potential professors can be a (small) problem. However, your professors might know some contact in a different research group. It might help you if your advisors contact their connections and make them aware that your application is in the application pool. This will prevent your application from being filtered out and your application might be more visible to other professors as well. Being admitted and selecting your advisor are often two different processes. So what you might want to do is to write your SOP to display interest in your old work combined with the work that you are interested now (combining ML with DNA testing sounds like this is actually the truth). Once you are admitted you can focus on finding an advisor that is willing to work on the new work, or alternatively you could be co-advised across domains.
I think your application is pretty strong even if you do not have those direct links though. I think it is all about that your application comes into the right hands. The right person should immediately recognize your research potential. But you might be unlucky that this does not happen. I think you have a good shot at the universities that you mention, but also make sure to send out around 10 applications with some universities as safety options. Good luck!
Rohit says
Hi Tim,
Really appreciate you replying to all the comments here. Very informative.
I have an undergrad degree(in CS) from one of the top 10 universities of India with GPA of 8.6/10.0 and good GRE and TOEFL score. Published a paper in IEEE International Conference on Big Data (Data mining related). I have strong recommendations (None of them know any professors in US). I have internships in top companies and 2 year work experience as core ML Engineer in a startup. I am not sure of the research topic yet (it will be among reinforcement learning or NLP).
Do you think I will have any chance of getting PhD admission in top 20 US universities?
I am also considering Masters and then hoping to transfer to PhD program. Is it a viable option?
What are the chances of getting Masters in top 20 US universities?
MS in DS vs CS – Does it matter? My observation is that getting into MS in DS is easier than CS but does it affect my further PhD application?
I know it is too much to ask. Can you please suggest some realistic universities for my profile?
Tim Dettmers says
Hi Rohit, it is okay if you do not know your research topic yet. Many PhD students don’t know when they start and become less sure when they discuss it with their advisors and peers.
I think you have very good chances of getting accepted at the Top 20 universities (+40%). Top 3 to 5 will be difficult. For the top 10 universities I would give you around +25%.
A masters at Top 20 US university is a very good option if you want to improve your chances considerably to be admitted to the Top 10 universities. I believe people value a more diverse profile a bit more than using BS/MS in the same topic. However, this has almost no weight on the overall admission decision, so do not make this any priority.
It is sad, but I would recommend you to apply to universities that do not have a racial bias against Indians. CMU, for example, is a university that gives any student independent of nationality or disability a fair shot — I recommend you to apply both to their Masters and PhD programs. However, some universities, like many Ivy League schools, have a racial bias against Indians. You can probably get a proxy of racial bias by looking at the proportion of Indian PhD students at the respective schools.
Nicole says
Hi Tim,
Thank you for writing this. I am wondering if you could give me some feedback on my chances of admission at UC Berkeley and CMU. I am an undergrad at UCLA majoring in Chemical Engineering. I have a GPA of 3.459/4.0 and 1.5 years of research experience. I will have 2 good letters of recommendation and 1 average one. I have a second author paper submitted (not yet approved). From your experience, would I have a chance?
Tim Dettmers says
Hi Nicole, I think it could work for CMU quite well. I am not entirely sure because I would need to see a bit more data, but probably have a good 35-65% chance of getting into CMU. One of the main determinants would be if your letter writers know some faculty at CMU. Unfortunately, for Berkeley, your chances would be much lower. However, again if your advisors have personal contact with people at Berkeley it could work out. I would definitely give it a shot! But also make sure to apply to many other universities.
Eric says
Hi Tim,
I’m a current EECS freshman at Berkeley and am pretty sure I want to go onto a PhD program for ML/AI (preferably top 10 school). I’m currently working on computer vision in a physics lab under a grad student, but it doesn’t feel very research-y and I feel like I’m doing a lot of implementation. I also probably won’t be moving towards papers or anything like that, as I’m sort of grunt labor :P. I was wondering if having research under a physics lab instead of EECS would be less valuable, and what I should focus on for the next 3 years.
I will probably look to join a different lab by spring semester of sophomore year, but research at Berkeley is very competitive (especially under professors such as Pieter Abbeel). I was also wondering if you had any advice on how to make the transition from working under grad students to having a personal project and moving it towards publication. For example, how should I approach professors with this intention without sounding rude or arrogant? Should I have personal projects in the field and a clear research direction before approaching them?
Thanks so much! I really appreciate it.
– Eric
Tim Dettmers says
First of all, as a freshman you have plenty of time to move forward in research, so no need to rush. The work that you are doing can still be quite valuable for you even if it is implementation related as long as it has some research background (are you working on software that can be used in research?). Since most undergrad projects usually do not end in a publication your advantage of moving to a different lab are not too great. However, the recommendation letter of the professors would be important. In most labs in Berkeley you will not have direct interaction with a professor and so the grad student will relay that information to the professor for a recommendation letter. The most important question is here: Under the guidance of your grad student, do you gain “visible” research experience, which can be put into a recommendation letter? Writing research software and contributing to a research project of the grad student would be “visible” research. Switching labs would make mostly sense if you would earn much more research experience in the other lab and earn very little experience right now.
The best way to connect to another lab is through grad students. Sometimes you can connect with professors directly, but this high depends on individual professors. If you know Pieter Abbeel talked to other undergrads directly, then you can try too. If he usually does not talk to undergrads about research then try to get in touch with one of this postdocs or PhD students. Same rule goes for postdocs though: If they usually do not talk to undergrads you should not try to get in touch like that. From what I heard, especially at Berkeley, there often can be a hierarchy where professors mostly interact with postdocs, postdocs interact with PhD students, and PhD students interact with master/bachelor students. Try to figure out what the hierarchy looks like in labs you are interested in, and get in touch in the appropriate way. Often it can be appropriate to email professors and ask them to redirect you to a relevant PhD student, but not always do professors reply to such emails. The best way is probably to get in touch with other undergrads working in a lab and ask them how they got started and then do the same.
Jim Mirzakhalov says
Hello Tim!
Thank you so much for the blog post. It was really helpful to have a deeper look into the pool of applicants and potential things to look out for. I am applying to PhD programs in the next month or so, and my area of interest would be NLP. I would like to assess my chances of getting into schools like MIT, UPenn, UW, UMD.
I go to a not well-known school in the US (University of South Florida ~100th). I have been involved with research since my sophomore year, and will have at least 2 strong recommendation letters from two of my research mentors. In terms of publications, I have 1 Deep Learning paper in review for IEEE Transactions on Mobile Computing (4th author), and another two HCI papers in review for a relatively top HCI conference (1st author in both). I also have 2 provisional patents filed by the university. Over the last summer, I interned at IBM Research, where I worked on a somewhat unrelated topic (cloud computing research) and we will be filing for a patent for the work, but I don’t think it will happen until I apply. One of my recommenders is my mentor from IBM. I just took my GRE (157 V + 163 Q) and my GPA is 3.84.
I don’t know if that matters at all, but I have been very active on campus by founding a relatively successful CS student society and have won a few hackathons.
I am very worried about the fact that my interest in NLP started quite recently, and I don’t have any published work in the field. What do you think my chances are for getting into those programs above specifically for NLP?
Tim Dettmers says
I think your chances are quite good. You have a strong profile in HCI and great recommendation letters in general. That the industry internship was in cloud computing research is not negative and it is a great bonus if you are filing for a patent. Make sure your mentor mentions this in the recommendation letter. The difficulty, as you already mentioned, is to convince NLP faculty to take you on. Most academics would value a research background of any kind (more applied researchers also value patents), but in NLP it can be quite competitive and some academics just want to see NLP publications. Your statement of purpose would be really important in this case. You might be able to get NLP people interested if you position yourself between NLP and HCI, but you should also convey your honest research interests. Don’t worry, the statement of purpose is not binding and you can negotiate the research that you would be doing with potential advisors during the visiting days — the statement of purpose can also be seen has showing off how you think about particular research direction and how it is related to your research experiences (rather than what you really want to do at the university). If you can tell a compelling story in your statement of purpose you probably have good chances for UMD. UPenn and UW will be a bit more difficult. MIT is very difficult, but could actually work out, because they often value a background that is hacky — definitely expand on your hackathon experiences and how they tie together with your research in your MIT application. Also mention your hackathon experiences for other applications, but keep them a bit more in the background. Your leadership experience is value for any application. Make sure it is well visible in your CV.
Overall, I would recommend you to apply to a couple more universities. Make it 8-10. Your chances for UMD should be +75% and for UW and UPenn +50%, for MIT maybe about 20%. With a couple more universities you have great chances to be admitted to a couple of great universities!
adas says
Hey Tim, how difficult is it to get into the Allen AI2 PYI program specifically for the PRIOR (Vision) group? Does research experience at university labs suffice to get into this program or does one need a stellar profile with multiple publications to get in like with other AI residencies?
Vin says
Hey Tim,
I’ll be applying to PhD in ML this fall and wanted to get your feedback about my chances to top 10-20 schools. I am a CS and Chemical Engineering double major at UMich (top 20 in CS). My GRE and GPA are fine and I have 1 very strong letter of recommendation. My only problem is my relative lack of CS/ML research experience. I spent much of my college time trying to figure out what interested me, and it took quite sometime before I could settle down to CS and ML specifically (although I am now confident that this is where my interests lie).
That said, I have done some research in Chemical Engineering and have a first-author publication in a peer-reviewed journal (similar impact factor to JMLR although I understand its tough to compare between fields). I’m also currently working (started this summer) on a first-author publication with an ML professor, although that work will probably not be submitted until after PhD Apps are due (but hopefully before!).
Also, I’ve applied to NSF, so if I do somehow win the grant, what do you think my chances are then?
Tim Dettmers says
If you already have one strong letter and you are getting another one from the ML professor you are currently working with your chances should be pretty good. The chemical engineering publication will look good but it will not convince all committees. A preprint might be helpful, but non peer-reviewed publications are not valued too highly — the recommendation letter from your ML professor is valued much more highly. In cases like these, your GPA is highly valued too. If your GPA is close to 4.0 this would increase your chances greatly. I would say for top 20 schools you would have about 30% per application for a GPA around 3.7 and 45% for a GPA around 4.0. If you apply to 10 schools you will have a good chance of getting accepted at some schools. In your case, you should also try to have 1-3 fail-safe universities which are not great but which you would be happy to go to anyway. In this way, you minimize risk at the expense of better possibilities and you should strike a trade-off with which you are happy. You can also always apply to more universities but this costs more time and money. If you get an NSF fellowship that can be a bonus, but most other factors are more important.
You could also try to apply for some internships which help you to bolster your profile for application in the next year. The predoctoral young investigator program at AI2 might be great for you. Residency programs would also be great, but they are also very competitive and might be out of reach for you. Another good way to get an internship is to ask your ML professor if he/she has some personal contacts in ML where you could do an academic internship for a year or so.
So overall your profile looks quite good. With a bit of luck, you can make it some very good schools so definitely apply! If you invest another year, you can make it easily to most top 20 schools. You can explore both path at once and decide based on what you get. Good luck!
adas says
Hey Tim,
What is your opinion of the AI2 Predoctoral Young Investigator Program . Does spending a year or two there boost one’s chances of getting an admit from the top 5 schools?
Tim Dettmers says
If you want to improve your chances of getting admitted this is a perfect program. It also helps you to understand if you like research and in particular research in industry (AI2 is actually more like between industry and academia due to its non-profit status). Based on your experience you can apply to PhD or maybe you changed your mind and you will be in a good position to find great jobs in industry.
bing says
Hi Tim, Thanks a lot for your great post. It helps a lot when I prepare for the applying materials. I’m applying for a PhD in data mining this fall. Hope I can get some advice from you. I have already talked with some professors at Purdue and Virginia through personal connections. However, I didn’t receive many positive applies (only two or three )from professors in other universities.
My background: 101(29+28+21+23) 319(151+168) I worked as a research assistant on information retrieval at top2 internet company in mainland China for the past one year. I got my master degree majoring in CS in hongkong with a scholarship (top2) and my GPA is B+/A+. I also worked as a RA for several months in the school lab. Besides I got my bachelor degree majoring in EE at top3 universities in mainland China and my GPA is 3.8/4.0. As for the recommendation letter, I have 1 strong letter from RA supervisor, 1 letter from msc dissertation supervisor and one strong letter from my bachelor dissertation supervisor.
In the following limited time, should I take Toefl again to get higher oral scores for TA, or to improve GRE? My target school includes UIUC, UCSD and UCSB. What would my chances be for Top 20 university? I’m having trouble determining what schools are a reach due to I didn’t have publications. Thank you very much for your time and effort!
Tim Dettmers says
I think your profile looks pretty good! Often recommendation letters are more important than mere publications and give you a boost. Your GRE looks great and I would not worry about that. Your TOEFL also looks good. It is only important that you reach the minimum levels of TOEFL for the universities that you want to apply (usually around the 18 range for speaking/writing). So I do not think you need to retake them.
If you apply for Top 20 schools I would say your chance of getting admitted per school are around 30-60%; I think for UIUC, UCSD and UCSB it should be around the 60% mark. I would not apply for the very top schools (MIT/Stanford/Berkeley) and focus your time and effort on other schools. If you apply for about 10 schools you should be able to pick from 2-5 schools.
If you want to improve your chances you can ask your contacts from Purdue and Virginia if you can intern with them. If you can manage to do that for a year and are able to get a publication under your belt then your chance should increase further to +75% and the top schools come into range also (20-30% for Top 3).
Hope this helps — good luck!
catcher says
Hi Tim,
Thank you very much for your detailed description of the whole PhD app process! Really I’m applying for PhD/MS in ML/RL (reinforcement learning) this fall. Hope I can get some advise from you.
My background: UCLA undergraduate. double major (cs+math), 3.8/4.0, 324/340, 2 second-author AAAI in RL submitted (decision is not out, but will be presented in NeurIPS workshop), 1 first-author small conference in RL submitted (decision is not out), 4 second-author SCI material science journals in application of ML, 2 fourth-author paper in adversarial attack, also working on a top RL conference (AAMAS) first-author short paper.
As for recommendation letter, I have 1 strong letter from a material science assistant professor. For the other 2 recommendation letters, 1 is from a AI/CV expert (I did the AAAI RL works and a graphical learning c++ library in his lab), the other one is from an assistant prof working on RL (I did the two first-author RL works in his lab). But the situation is a bit interesting here. They are pretty busy (consistently flying all over the world), and trust my research and writing, so I have to write my recom. letter drafts.
I’m not sure about the weights of non-first authorship. I’m wondering if I should apply for PhD programs or MS for top schools (eg. top 4 + uw + cornell + princeton + harvard…)? I heard Berkeley MS (not MEng) is even hard to apply than PhD in CS. Is that true? Also, more importantly, what should be the focuses of my letter drafts separately? And what are the caveats?
Thank you very much for your time and effort!
Tim Dettmers says
What I have seen that some universities do not care if you did research in another domain (material science). Other see it as research maturity and see you as a highly valuable candidate and must have. Even without the material science papers and all papers being rejected your profile is still very strong due to your research experience. If some papers make it in, I can easily see you being accepted anywhere where your material science background is valued and you will have good chance for any Top 5-20 program.
Writing your recommendation letters yourself is not unusual. What your professors usually do is to delete a bit here and there and then add a personal paragraph and then submit the letter. Your letters seems to be quite strong.
Definitely apply for PhD programs. Your chances are very good for UW, Cornell, Princeton, and Harvard (+66% accept). You have also good chances to get into Berkeley, Stanford, and MIT (+30%) so definitely apply for those too if you can find interesting advisors at those universities.
I am not sure about the MS application process, but it is also a great option if you can afford it. MS students from a university are very often accepted to the PhD program at the same university. Additionally, Stanford and Berkeley are usually very happy to swap MS students for their PhD programs.
Let me know if you have more questions!
sjs95 says
Hey Tim,
Massive thanks for this post.
I graduated from USC with an MSEE in with a 3.8 GPA. I have 2 publications in IROS, 1 in ISBI (top3 biomedical imaging), and 1 submission in ICLR. My recommenders don’t have any personal connections but they actively publish in their field every year and are known for their work. But I am the second author on all the above-mentioned papers. What are my chances at CMU, UCB, Stanford and UWash?
If I fail to get an admit this year I am also applying to various AI residency programs.How much will that improve my chances?
Tim Dettmers says
I think your changes are pretty good especially for CMU and University of Washington. Stanford and Berkeley are more difficult to get into and I would see your current profile to give you 30-40% chance for either university. So if you apply now you have a good chance of making it into some good universities and with some luck into Stanford and Berkeley. I highly recommend applying for AI residency programs. It gives you more experience and also gives you a better outlook into what do you want to do with your career. If you get into a residency and a good school, I would accept the residency and defer the school entrance to the next year. If you do not get into the schools that you want but into a residency program, definitely do the residency program and just reapply next year.
Jason says
I go to Dartmouth College and i have a 3.9 gpa. I have one EMNLP publication, one neurips workshop publication, and four medical journal publications on DL for medical image analysis. My advisors so far dont have any connections.
Do I have a chance at a top NLP PhD program or should i wait a year?
Tim Dettmers says
Yes, you will have great chances to get into top NLP PhD programs. I would apply this year!
Cybernetic Pupper says
I study at a country where undergraduate research is very uncommon. However, I was accepted into an internship abroad and I will (hopefully) get one or two top conference publications (as first-author) out of it, plus one more top conference publication but as second or third author, however they will not published in time for the December 15 applications (the conference decisions will come later than that).
I have a good GPA. I will only have research recommendation letters from my supervisor at the internship, the two other letters will be from people who knew me only in class.
Should I bother applying anywhere? Do I have a nonzero chance? I don’t necessarily want top schools. Reading this post is honestly very depressing and makes me feel like I should just quit this idea of getting a PhD in the US.
Tim Dettmers says
Sorry for making it seem a bit depressing. Many people want to go to the very top schools and I wanted to give them realistic feedback if it is possible for them. If you do not necessarily want to do a PhD at a top school then please apply — if you get more than one research paper before the application deadline it is already very impressive. One good research letter is also great. I personally had one great research letter, a letter from someone that I took a class with, and a letter from industry. Such a mix of letters could also help you. However, if you still feel your profile is lacking, this is also something that you can address at the end of your statement of purpose. In my SoP, I mentioned that I was unable to secure other good research letters because I was just not in the right environment to get those letters. If you also feel that others had more privilege than you and you just not had the right opportunities then this might worth adding. I think overall you have great chances of getting accepted by most Top 10-20 universities (or below) in the US. If your advisor has some connections with some universities, you should have a very high chance of getting accepted. So do not despair, I think you got a good shot at this!
Zach says
Thanks for the great post!
I’m applying to both PhD and master programs for 2020 fall. My target school includes UCSD and UCSB. My GPA is pretty low (around 3.44/4.0, 83/100), my GRE and TOEFL are 156+170+3.5 and 109 respectively. I only have one first-authored paper at ICLR 2019 conference (I couldn’t find any collaborator, I really wish I could, but my undergrad school wasn’t prestigious. I basically had to work by myself). My question is that would it be better if I first get a master degree in the US, and then decide whether to get a job in the industry or PhD depending on my research? And how likely I’m able to get an offer (Master or PhD) at schools like UCSD/UCSB/USC. Thank you!
Tim Dettmers says
I would apply for both Master and PhD at UCSD/UCSB/USC. I think you have not bad chances of getting accepted for both a master degree and PhD degree. If it does not work out a master degree or an internship would be the right thing if you want to get a PhD. If you can get another paper and recommendation letter out of a masters/internship you should have very high chances of getting into your selected schools. However, also apply more broadly. Try to pick 10 schools or so. You can always do your masters somewhere else and then try to get a PhD offer from UCSD/UCSB/USC.
Chris atkeson says
You might want to suggest that people apply to both masters and PhD programs at the top schools. Getting in to a masters program is usually a good way to get in to a PhD program. This is true at the CMU robotics institute for example.
Tim Dettmers says
That is very true! There are some other good strategies one can pursue but if one is applying for PhDs already, this is the obvious one. I should add it!
Willie McClinton says
Hello Tim, Thanks for your post. I enjoyed how it gave some estimates of what credentials students are applying with now-a-days. The competition is getting rough, but all for the better. I’ll be applying for PhD programs next year from my undergrad institution and I was wondering if you mind giving me some advice on what schools should be on my list to apply for? My goal, like many others, is to get into a program like MIT, Stanford, or UC Berk and my interested are Deep Learning and Reinforcement Learning.
I am getting a degree in CS with minor in Math from a not well known US state school (US ranking = ~100th), but I have been actively involved in research since freshman year leading to a few publications : 1 for research done during my Freshman year summer at NIST (At a not well conference), 2 for research done in Brain-Computer Interfaces and Human-Computer Interaction (HCI) at my school (1 an oral at a Tier B HCI conference and 1 a poster at a premiere HCI conference; both full papers published in proceedings), and 1 possible IEEE journal paper being drafted in Compute Vision + Deep Learning. Also, this summer I am doing research at top 20 university in Deep Reinforcement Learning with a well-known advisor in the field (many connections) and we are trying to hit the ICLR deadline in September, so I can actually have a top publication in the field I want study before grad school applications.
I haven’t taken the GRE yet, but I’m pretty confident that I’ll do fairly well. I have a 4.0 GPA, won a Goldwater Scholarship for research ability, and have a couple of cool open source projects because I compete at a lot of hackathons (some have won prizes).
I was wondering how heavily does not attending a prestigious institution for undergrad affect my grad applications and, with that, what will my chances be for getting into a program like MIT, Stanford, UC Berk, and what would my chances be for top 20 university? I’m having trouble determining what schools are a reach due to my undergrad institution not being well known. Also, what are some things I can do within the next few months before applications in the late fall to improve my chances for grad school? Thanks for your time and help.
Tim Dettmers says
Your profile is very strong and for many top 20 schools you should have no problem getting accepted. Especially universities like CMU and University of Washington would favor the profile that you have (+80% accept). You probably have good chances for MIT as well (+60%). For Stanford and Berkeley you will have moderate chances — I guess about 30-40% each. Looking very good — keep up the good work!
Evgenii says
Hi Tim!
May I ask for your advice regarding my situation?
Currently I am finishing my MSc in Statistics and Data Science at the University of Edinburgh (half of my credits are Statistics/OR and the other half is ML related). My current GPA is 81/100 (I have all A’s for all courses) which translates into 4.0/4.0 in the US system. I was conditionally accepted to PhD in Biomedical AI in Edinburgh but this may not happen as funding situation is not that easy for non-EU students (so I may have to do PhD elsewhere, e.g. US/Canada). My research thesis is related to statistical genomics and may or may not turn into a paper (depends on the outcome of research and time allocated, as not a lot can be done during 3M project). Some of Edinburgh’s famous people in ML promised me to write references in addition to my supervisor (who is statistician). If Edinburgh is off the list, I will be doing research internship with a professor in Data Science group in USC during few months of 2019-2020 academic year (which again may turn or not into a paper), but will still count as research experience.
Oh, and if that matters, I have a 4-year working experience in World Bank in Washington, DC in Quantitative Analysis Team.
Do you think I have good chances of getting into CMU for PhD in joint Stats/ML or say NYU for PhD in Data Science or should I not aim that high? I am not from a traditional CS/Physics background, I did Economics/Mathematics as undergrad but I do have a solid math knowledge and pretty good programming experience (Python,R,C++,VB,a bit of Shell Scripting).
Regards,
Evgenii
Tim Dettmers says
I think the USC internship will be the tipping point for your application to go from good to strong. If you work well at USC and are able to get a good letter from your advisor there you will have strong letters and a good range of research and work experiences. I think in particular for NYU you might have very good chances (>40%); CMU can be a bit tricky but they also usually also value more diverse experience and I would guess your chances are >30%. I think if you apply broadly you can be very sure that you will be accepted at a very good program. To be accepted at your dream program is always a roll-of-a-dice but it also looks pretty good for you. Good luck!
Michael says
Machine learning and computer vision are from what I hear very math intensive. What math courses beyond calculus 1-3 should a undergrad take to prepare for a PhD in computer vision?
Tim Dettmers says
If you did calculus 3 you are ready to go. Any additional knowledge is usually learned as you go. Learning something before you need it is not efficient and will slow you down. If you want to prepare for a PhD I recommend (1) doing a research internship, or (2) try to work on a research problem, or (3) find research problems that might be interesting to you and your advisor. Having better math skills will have little effect on your performance in your PhD.
Lothar Budike says
I recently read your paper on The Brain VS Deep Learning Part 1 and you made a very good case on how the Brain is not completely understood as to its learning capability by humans at this point and to compare it with Deep Learning, the technology does not even exist yet for machines to do computational analysis due to bandwidth, power, memory limitations etc. I was wondering, you are assuming millions of computers being networked in neural configurations against 1 brain. Have you ever considered that around 70K people show IQ’s above 156 and if they were all working in concert on the analysis it would always be ahead of the computer? Why one brain and not a collective effort on the human side? Lou
Tim Dettmers says
This is a good point and it shows that individual smarts also have their limitations. Neanderthals had larger brains than we do but they had less sophisticated social structures across tribes and because they had less knowledge-transfer, collaboration and trade they might have met their demise while humans flourished. So yes, I think this is an important point on top of all of this. We would need not only computers as fast as the brain but computers which can collaborate (or just one very, very big computer).
Matthew says
Hi Tim, great post. I have been admitted by some universities now. Could you please give me some advice which one is better for my further phd application?
1.msc machine learning,imperical college london
2.msc big data technology,Hong Kong University of Science and Technology
3.master of AI, University of Lugano
4.master of CS, Univeristy of Tokyo
It seems that msc degree in HK and UK does not provide abundant research experience and nearly no publication, not to mention top conference. Also , most foreign students(97%) participating in 2 years’ program at The University of Tokyo have no publication but only research experience.
master of AI in Lugano with IDSIA can provide enough opportunities focusing on AI. many famous scientist are working there such as Michael Bronstein, Piotr Didykbut and Schmidhuber, but this university is not so prestigious compared with HKUST, UT and ICL.
Could you give some advice on which one to choice if I wish to pursue my phd research in UK(ICL UCL) or Switzerland(EPFL ETH)? Really appreciate your reply!
Tim Dettmers says
This is a tough choice. I think to make it work you definitely need to spend your summers in research internships and you should get involved with research early in your master. I studied in Lugano and it is not necessarily easy to get involved in research. If you go to Imperial College I think you might be able to (1) do research there, or (2) use your connections from imperial college to intern at UCL/Oxford/Cambridge/Microsoft Research/ICL in the summer (you will not have many similar connections if you go to Lugano). You should write some students in research labs and ask them if they usually have interns/student workers at imperial college. Otherwise, there is always (3) do a year of research internships after your master. I also did this and this is a rather save path to a good PhD program. You will need to use your connections that you found during your master program to find these after-master-research-internships. Good luck!
Nitin says
Hi Tim,
Thanks for this great post that I just came across! I completed my Bachelors in Electrical Engg. from a top 10 school in an international country., and after that I worked as a software engineer where I did basic NLP/ML work, nothing research-y.
I got into a US school ranked around 50 for MS in CS, and have done 2-3 research oriented projects in which one project led to submission at a journal (which can take ages to get published). I also had a research oriented internship at a startup working on computer vision.
I plan to apply for a PhD starting in Fall 2020, but honestly after reading your post, it looks kind of impossible for me to get into a top school for a PhD in AI, as I do not have any publications ( I don’t know how everyone can get 2+ papers). Is there anything I can do to boost my profile, or should I just settle for the university that I am in currently?
Tim Dettmers says
If you worked on some research projects then you probably worked with some researchers. You can ask them who they know and if they know someone in a top PhD school this might help your application. Alternatively, you can just do one year of research internships. I was in a similar position and I also did a year or research internships. It may be difficult to get a research internship (especially an academic one), but you can ask the people who you did research with if they know someone who would be willing to take you on.
Nitin says
Thanks for the response. Yes I have asked my professors and I may be able to continue to work as an RA for my current adviser for a year and then hopefully get a publication by the application season. Unfortunately, most research internships in industry require you return to school after interning with them, which is not possible in my case as I’m graduating.
Alex Veuthey says
I think the opposite can be said, regarding finding jobs with or without PhD in case of an AI bubble collapse.
If/when it happens, there should logically be less money invested in AI, and then recruiters might prefer candidates without PhD, who would claim lower salaries than candidates with PhD… Being overqualified can also be bad!
Tim Dettmers says
That is an interesting way to think about it and could well be true. I guess the people at the fringes, either cheap and qualified, or expensive and highly qualified could be high in demand.
Tayyab says
Tim great post.I have a question,i struggle too much at university level.Have a low GPA.
My university ranking is in top ten in country.Reason for low GPA is having ambloypia in
my left eye.No idea what to do,and how to deal with it.
Tim Dettmers says
It is not easy to do well with amblyopia! But remain steady — if you keep at you might be able to get into good universities. What you should do is try to extend the time that you spent on research and accept that you need a bit more time than others — for me, it was also like that! This could be for example that (1) you do a master degree (if you do not have already one), (2) try to do a gap year to do some research at some university, (3) do research internships after your bachelor/master degree. If you have a good research background, nobody will care about a low GPA and bad school! Also, read this.
However, you should also be realistic. Even if you spent more time than others on research, you still might not be able to get into top universities. But keep in mind that you can find excellent mentorship in many other universities in the top 100. Often, you can find an advisor which fits you perfectly at a top 100 university, and you even though it will not be as prestigious, you will get a great PhD education! If you need more advise, please feel free to email me.
Lucas says
Is the post about research internships coming up and/or do you have any quick tips? I’m starting to look around. The current plan is to email professors that do work that’s very interesting to me, plus google around for official programs + NSF REU listings. Though I’m extremely unfamiliar with the process so advice is appreciated.
(For context on me, I’ve interned a lot at top companies, only recently got into serious research and I’m in the process of putting together papers. Plan is to apply for grad school next year. I figured that doing visiting research would be great for exposure and to potentially get connected to a future advisor. Undergrad at top 10 school and above 3.5 GPA if it matters.)
Tim Dettmers says
Hi Lucas, that is a good plan. The best way to find research internships at universities is to try to ask your connections if they know someone who does research that could be interesting to you. These people can connect you and often you can get a research internship that way. Otherwise, you have to go through official channels. It might also be possible to get an internship by first going through
the PhD students of an adviser that you are interested in, but this is always a bit tricky and it will probably unsuccessful most of the time. So your best bet is really to look at what connections do you already have.
Harry says
Hi Tim, Thanks for your great post. It is really helpful. I am going to apply for phd next year. Could you please give me some advice on my situation ？My goal is top 30 university for America.
I just got my master degree majoring in AI at the University of Edinburgh. I got my bachelor at Nankai University (top15 in China). I am going to work as a research intern at Czech Technical University in Prague for 1 year (start from next month). I got 101(28+29+20+24) in Toefl and 321 (151+170)+3.5 in Gre. I do not have a lot of official research experience and publication now.(My master thesis advisor says that she is going to publish a paper based on my thesis. My name will be included. But it is not published now). My research interest is algorithmic game theory and reinforcement learning. I do not care about the country to pursue a phd a lot, I just want to go to the best university based on my background.
Do I need to take another toefl or gre test? I think I can get better mark, if I work hard on it. However, based on your post, I think it might be better to focus on my research in Czech Republic. By the way, my GPA is not good(3.3 undergraduate, 3.6 master).
Tim Dettmers says
Hi Harry, your GPA and GRE are borderline, but still okay. I would not worry too much about that. If you want to get into a top program I would try to do excellent research at your position in the Czech Republic and try to publish papers at top conferences. If you do not have any papers in the next six months, I would suggest that you try to do research at another institution for another year. This would give you a strong research background and I think you would have a much easier time to get accepted at top universities. Good luck!
Bob says
I don’t agree “Don’t contact advisor before application”. Many students I know get their offer by introducing themselves to advisors. I thought your options fit to TOP 5 graduate school since the committee will make all the decisions. Could you explain why you support not contacting advisor before application?
Tim Dettmers says
It always makes sense if your advisor has a personal connection to faculty that he/she introduces you to a potential advisor. Contacting advisors personally can make sense in your case since you have a strong profile. Trying to contact faculty personally if you have a weak profile can be a waste of time and in the worst case can be detrimental. In your case, I would go for it.
Bob says
Hi Tim, thanks for your very nice post, really helps a lot. Could you help me analyze my situations? I will start application next year. Currently I am a junior in a university in Minnesota ranked around 30 in csrankings and working with a faculty. Now I have 2 papers in top conference of computer vision and robotics and next year maybe I will have 2 more. For the recommend letters, there are two Professors(an ieee fellow and my mentor in lab) who knows me well and can give me strong recommend. My GPA is not good, just around 3.5. I was wondering what should I do in next year so that I can 1, cover the short come of my undergrad weakness and low GPA and 2, strengthen my competitiveness? Thank you.
Tim Dettmers says
Your profile is strong. GPA does not matter in your case, especially if you manage to have 4 publications by next year. It can even be an advantage which shows that you care more about research than classes. Make sure though, that your GPA stays above 3.5.
For your letter writers, it matters a lot if they know faculty at the institutions which you want to apply to. Ask them who they know.
To improve your profile make sure that you have a good third letter. This could be done with a summer internship for example. If you want to make an overwhelmingly strong case for top universities, I could recommend residency programs. These programs yield experience which helps you later to decide if you want to go into academia or industry. Additionally, you get another good letter and possibly 1-2 publications.
With 4 publications and if your letter writers have personal connections I think you have about 30% admittance probability for the top 4 schools. With a residency program, you can push this easily to 50-60%.
dheera says
I got into MIT, Caltech, and Stanford. I had exactly 0 papers in “top conferences”. The vast majority of my classmates at MIT did not have any conference papers coming in. I highly doubt the 93% statistic.
That said, I *did* have research experience at multiple internships and in school and I was very proactive in building tons of stuff in my spare time.
And if you have even 1 paper in a top conference that’s a huge plus. 4 papers in top conferences is honestly more like what you have when graduating from a PhD program.
Tim Dettmers says
Thanks for your feedback and your example. I think I just messed up with presenting the statistics in an understandable way — I will rework this later today. So actually these statistics also cover your case (they are just presented in a way that makes this difficult to understand). For example, if you did your undergrad at MIT, Stanford, Berkeley, CMU, you need fewer publications to enter (usually between 1 and 2 for the school above). However, if you did not attend a top 20 university or the best university in your country then you will need 3+ publications to enter (the average was 6 publications for people who did not do undergrad at top schools). I also know that MIT does not select as aggressively for undergrad degree, so your example is more common at MIT (especially if you hack some projects together — MIT loves that!).
One additional note: In current times and especially in the field of AI we live under publication inflation — too many applicants have one or multiple publications. One publication to enter the PhD was a lot 10 years ago, but it is not any longer, especially in fields like deep learning. I know some faculty who made it their rule to not take any student who has not published at a top conference before. That is harsh, but that is just a reality at the moment. In the end, its all about supply and demand — and the supply is increasing while the demand is steady.
Daniel says
Very nice post but it does seem to that a lot of this applies more particularly in the US. At least my experience when applying for a PhD in Europe was vastly different. Getting in touch with potential supervisors, interviewing and writing a research proposal was a very large part of the application process while the weight of the statement of purpose was I think way lower. I didn’t need the GRE either. It probably also depends where in Europe but this is the feeling I had with both the UK and Switzerland.
Tim Dettmers says
You are right, this post is mostly about applying to the US and experience will differ from PhD applications in Europe. However, I also see the research proposal in Europe a bit like a drawn-out statement of purpose, where the focus is on future research (and a plan of it) rather than what research one has done before.
su says
If my TOEFL score is below seven points from minimum, will I be rejected even in the filtering process?
Thank you so much!
Tim Dettmers says
For many university, you might be filtered out — sorry! However, some universities are not as strict.
What might help is if your recommendation letter writers know someone in the university this person might be able to save your application from the filtering process so that it is at least looked at. You can also tell your recommendation letter writer about the situation and your recommender might mention this in the recommendation letter: “X scored below your minimum TOEFL requirements, but I believe that X’s command of English is sufficient to work in a team and publish research”.
Tommy says
Tim, this guide is excellent. Thanks for publishing!
Michael says
Wow, 98% applicants have 3+ publications in top conferences? That’s pretty tough!
Tim Dettmers says
Well, some might have only two publications but they went to a top school and had strong recommendation letters and so forth. But yes, that is pretty tough!
Michael says
Why did you choose UW?
Tim Dettmers says
I might write about how to select among your offers in another blog post. For me, the main criterion was a combination of personal reasons and a good fit with potential advisors at UW.