Questions that I had


What does a strong application look like?

Here's what I think is the most important thing to understand about your application: although you are technically applying to schools, you are actually applying to professors. Whether or not you ultimately get admitted will likely come down to just one or two of them deciding that they'd be happy to supervise you. A strong application is therefore one that speaks directly to the professors that you are interested in working with. To do this effectively, you need to have compelling answers to three questions:

1. What do you want to work on? 
2. Who do you want to work with?
3. How have you prepared to conduct this work?

Your answer to the first question is by far the most important because it should be the very reason why you are applying for a PhD. Throughout your statement of purpose and during your interviews, it should be clear that there are open research questions that fascinate you and that you are willing to dedicate at least the next five years of your life to answering. It is good to have a maximally concrete understanding of what these questions are, but you don't need to have specific ideas for papers just yet. You just need to be able to describe high level problems that you are genuinely excited to work on, and to have some rudimentary ideas for how you might go about achieving this.

Your answer to the second question is important because it should be the very reason why you are applying for a PhD at this particular department. And that is, because it happens to employ at least one professor who is interested in the same problems as you are. If you answer the first question with enough detail then it should be pretty obvious who these people are -- to you, to them, and to any other person in your field. Thus, answering this question well is mostly a matter of further developing your answer to the first question. After that, you just need to find the people who can best support your interests and identify general directions that you might both be interested in.

Your answer to the final question should ideally be that you have already begun to pursue the type of research that you wish to conduct during your PhD. You should have one or more research experiences that you can discuss, and great letters of recommendation from your supervisors for these experiences. Ideally, at least one of your letter writers is known and respected by the professors that you are applying to. In the absence of these things, you should at least be able to demonstrate that you are well-read and passionate about the area that you hope to conduct research in, and that you have given your best shot at cultivating the necessary background.

This all means that a strong application for one school might be a pretty weak application for another. It simply depends on your fit with the people there. What I found while applying was that the process was extremely smooth at the schools where the professors and I had interests that precisely aligned, but that my applications weren't as well received at schools where I felt like I had to push my research statement to fit their specific direction.


How do I get strong letters of recommendation?

The strongest letters are written by researchers who have directly worked with you, who have a good impression of you, who work on research that is related to what you are pursuing, and who are known and respected by the professors that you are applying to. Obtaining a strong letter of recommendation, and let alone three of them, is thus a multi-year effort. Long before the time comes for you to put your application together, you need to be planting the seeds of this effort by pursuing and excelling at undergraduate research opportunities, in part so that you will have good letters of recommendation when you later need them.

The good thing is that this is something that you should be doing (and should want to be doing) anyways. There's a universe out there where letters of recommendation aren't needed for PhD applications, but I am confident that its slightly less anxious inhabitants are still doing the exact same thing as what you should be doing: working on research that you care about with researchers that you respect, and in doing so discovering new streams of research that excite you deeply. If you can do that, then the letter will work itself out when the time comes. It will be straightforward for you to ask for one, and it will be straightforward for your supervisor to write one.

Getting a strong letter thus starts with figuring out how to get started on doing research in the first place. This can be challenging as an undergraduate, but the opportunities are almost certainly there. Look for summer programs, research subjects, thesis opportunities, internships, research assistantships and hiring opportunities for lab managers. Take a professor's course, do well in it, and go ask them if they'd be willing to supervise you on a project. Volunteer at a lab and offer to do free grunt work! Look for opportunities outside of your own university. Don't be timid, and don't be put off by rejection or failure. All of my current opportunities have stemmed from just one or two times when I nervously asked someone that I didn't know if I could work with them, and there have also been many times where I had to backtrack a little after starting something that didn't work out.


How many publications do I need?

Zero is fine, but ideally at least one. Beyond that, quality and relevance are more important than quantity. I doubt that there is a single program where you need publications to be admitted, but what you do need is research experience that allows you to compellingly answer the above questions. Admittedly this is probably easier if you have at least one experience that resulted in a publication (or one in progress).

When I was applying, I had the impression that everyone who got into the most competitive programs already had many publications under their belt. In retrospect, there's clearly some sampling bias going on here: if you're searching online for student profiles, you're more likely to find the profiles of the students with the most publications. That doesn't mean that you need to have a similar profile. Anecdotally speaking, many of the admits and current students that I have met did not seem to have had publications at journals or major conferences at the time that they were admitted. It seems like it's more than enough to have a cool unpublished undergrad thesis and a compelling vision for what you want to achieve. This might be less true for programs in fields like machine learning, but I still don't think that a lack of publications should deter anyone from applying.


How important are grades and rankings?

I was really worried about whether my grades were good enough to be competitive for PhD applications. I come from Australia, where grades and rankings are weighted relatively highly for university admissions and scholarship applications (eg. undergraduate admissions here are entirely based on your statewide academic ranking, and I have the strong impression that PhD scholarships also follow this mould). In this system I was good but not great; I had a WAM (weighted average mark out of 100) of 84 and I never made the Dean's List (top 3% of undergraduates). This means that every year there were (at least) several hundred students in my degree cohort who were higher ranked than me, and several dozen students in each of my majors. I remember telling a mentor that I was also interested in coursework based Masters programs because of this (I wanted 'academic redemption', as I put it), and he basically told me that that was a silly idea. I now agree.

Getting good grades is important -- it demonstrates that you are familiar with your course materials, and that you have a certain level of discipline and intelligence. But for most people, getting top grades requires one to invest in specific test taking skills and strategies that are not particularly useful outside of the classroom. If this is the case for you, as it is for me, then time spent on this long tail effort of perfecting your GPA is probably better spent on endeavours like research, internships or side projects. You will learn and grow more, and that's what actually matters. Of course, if you can get top grades without sacrificing other experiences, then more power to you!


Should I email professors before applying?

Before applying you should try to make sure that the professors that you are interested in are 1) accepting new students and 2) interested in supervising the type of projects that you are interested in. If this information isn't available online then emailing them is a really good idea. And if they specifically ask prospective applicants to reach out before applying, as some do, then you obviously should. But otherwise I doubt that emailing ahead of time will give you an advantage later on (very small sample size, but I was equally successful/unsuccessful at the places where I emailed vs didn't).


How do I email professors?

Polite, brief emails with high information density, clear questions and easy potential responses work best. When writing your email, just remember that its purpose is nothing more than to find out if it is even worth applying, ie. if the person that you're emailing is accepting new students and if they are interested in the same things as you are. Thus, it should probably only consist of a quick summary of your past and future interests followed by one or two yes/no questions. Their reply will probably only contain their yes/no answers. 


Should I sit the GRE?

I don't know if this is true for all programs, but my GRE scores were totally inconsequential for the ones that I applied to. When programs say that GRE scores are optional, they genuinely seem to be optional. Most programs in my cycle didn't even accept GRE scores, and at one of the few schools that did, I'm pretty sure that they received my score only after I had already been internally admitted. I actually kind of enjoyed studying for parts of the GRE, but it is also super expensive and time consuming so I would not recommend doing this unless it is compulsory for the programs that you are interested in. 


What are interviews like?

I interviewed virtually with eleven professors across five schools. I only met with each professor once. Each interview lasted for between 20 and 90 minutes, with 30 minute interviews being the most common. Overall, I found that the interviews were very laid back and surprisingly enjoyable. For the most part they are just casual conversations where you talk about your past work and future interests.

Almost all of the questions that I was asked were variations of the three that I have highlighted above. Common phrasings of these questions were straightforward -- things like 'describe a project that you are most proud of', or 'what kind of projects are you interested in working on next?', or 'why do you want to do a PhD?'. When the conversation flowed naturally to problems that we were both interested in, I occasionally also received more intellectual questions like 'what do you think is needed for AGI?', or 'what do you think about Sutton's bitter lesson?'. There was no pressure to have a 'correct' answer here; it really just felt like we were having a fun discussion.

There were a couple interview questions where I was caught off guard. One was 'what is it exactly that you want to work on?'. I struggled with the 'exact' bit. Another was 'describe the most interesting research paper that you have recently read'. I found it a little difficult to choose one on the spot. These blunders taught me two things. First, interviews can reflect both the personal style of a professor and their capacity to take on new students -- some professors are more intense than others, and some might only be looking for relatively independent students because their labs are already very large. Second, the best way to prepare for these interviews is really just to continue immersing yourself deeply in the research that you wish to pursue. If the fit is there, then the interview will work itself out.


Should I not tell anyone about my application because it will be super embarrassing when I inevitably and spectacularly fail?

Don't keep your application a secret!!! Applying to programs can be really scary and embarrassing because of the seemingly high prospect of failure, and I am definitely familiar with the temptation to be super cagey about it all. But, like most things, you don't actually have anything to lose by talking about it. My application benefited immensely from the feedback and advice of a handful of mentors, friends and family, including people who had no experience with academia or my specific fields. Even then their advice was invaluable, and I wouldn't have made it in without them.



Resources

Advice from...

Roman Feiman 

Lucy Lai

Vael Gates

Kate Petrova

Jamil Zaki

MIT EECS Faculty

Harvard PRO-TiP


Example SOPs

CS-SOP.org

Past NYU C&P SOPs


Github Repos

https://github.com/pliang279/awesome-phd-advice