Grad School Admissions

This post is about my experience with the Fall 2025 Computer Science PhD application cycle, primarily for US-based universities. I’m writing this mainly to document my own understanding of PhD applications and provide other useful references for future PhD applicants. Since there are already many blogs and write-ups on this subject, this page should serve more as a resource than a suggestion; I highly recommend the reader to learn from the best and read as much of this material as they can. While my focus will be on computer science research, I believe that the overall process remains similar for other domains.

Building a research profile and starting early

Think about why you want to do a PhD in the first place. This is an important step to the process, and I myself spent around 2 years thinking about this. Unlike undergrad or masters programs, one usually looks for a good research fit as a PhD applicant. Essentially, forget about university rankings, and think more about the lab you’re applying to. Inevitably, recognizing a lab which will be a good fit for you will require good domain knowledge.

To build a competitive profile (especially for domains like machine learning/artificial intelligence, where getting good PhD positions has become extremely competitive), you should start working early. If you’re interested in AI/ML research, start going through papers accepted to reputed venues. Use tools like alphaXiv to find new papers. Use csrankings to find researchers working in different areas of computer science (more on this later) and read their papers. Once you read enough papers, you’ll get a good idea of your interests and the kind of work you want to do. You’ll also learn a lot about how CS research looks like, how papers are written and overall what topics people are interested in. The next step will be getting research experience.

The best way to get research experience is to contact professors who work in your area of interest, and to work on long-term projects with them (work which eventually leads to good publications). If you’re not in school, research assistantships (aka RAships) are a way to go. In my case, I worked as an RA at the Learning and Networks Group (during my masters) and the Laboratory for Computational Social Systems (after completing my masters). My work spanned both theoretical areas in ML (online convex optimization) and applications (LLMs) and helped me prepare well for my PhD applications. Pre-doctoral research programs (also sometimes called AI residency programs) are another great way of getting this research experience. I’m aware of such programs at Google Research, Microsoft Research, and IBM. Here is a nice repository containing a list of AI residency programs. Alternatively, research SDE jobs are another great way of getting research experience. In fact, many open source orgs like HuggingFace (or, recently ggml-org) hire for research + development roles. I should point out that these jobs require a relatively high bar of ML + systems skills.

However, my take on these residency programs will be to consider them with a grain of salt. I’ve personally known and encountered cases of favoritism and highly unprofessional attitudes from some of these programs. In any case, once you’ve taken these steps and built a good research profile for yourself, you can think of taking the next step. Through your experience, try to get good quality first-author publications, and make sure you’re well aware of work in proximity to your research problems.

The timeline

For the Fall 2025 application cycle, most US universities had deadlines in the first two weeks of December 2024. Infact, most US universities also follow an April 15 Resolution which give applicants time till April 15 (of the given year) to consider their decisions for enrollment. Though this is not binding (and many universities didn’t follow it for the 2025 cycle), it roughly gives a good estimate of the overall timeline of the application process. After completing the application process, I started receiving invites for interviews during the early part of January 2025. All interviews were finished by the end of January 2025, and I got to know about my acceptances in early January and mid-February.

The application timelines for European universities vary a lot and depend on the advisor/research group as well. I’ve heard of people applying to PhD positions throughout the year. I applied to one European university (based in Netherlands), and the application process for that position was between January and February; I was told of my acceptance in mid-February.

The December deadline is for all components of the application (which I’ll discuss in a bit). My advise, as before, will be to start as soon as possible. A big part of the application involves a lot of writing; do not leave this to the very end of the timeline.

Application materials

Roughly, all applications have this format:

1. An application form (sometimes with questions phrased as prompts to be completed).
2. A statement of purpose (SOP).
3. A personal statement (not required for many universities).
4. A resume (CV).
5. Atleast three letters of recommendation (LORs).
6. TOEFL/IELTS scores (no GRE required).

A lot has been written about how to write the SOP and the how to build a good resume; it will be pointless for me re-iterate those points here. Instead, I again recommend the readers to learn from the best. For starters, go through this website to get a lot of examples of SOPs written by really talented applicants. Particularly, try to see the structure of the SOPs. You’ll notice that there’s no one template that works, and infact different applications have different writing styles. Still, having structure in your SOP makes it easy to follow. Checkout Stanford’s guidelines on writing nice SOPs. Additionally, check out the links I posted on the top of this page. For the personal statement, I found this example by Eugenie Y. Lai to be quite appealing. Also check out Berkeley’s guidelines for the personal statement.

Application support programs. Additionally, you can sign up for application support programs by various universities. These programs can sign you up with a grad student from a reputed university who will assist you in graduate applications, either by giving single/multiple rounds of feedback, or even full-fledged assistance in some cases. Check out the MIT GAAP, Stanford, Cornell and Princeton’s support programs. Also look up other support programs. As always, apply for them early (I unfortunately missed two deadlines that I wanted to apply to, due to laziness).

Finding potential advisors

Now this is the most important part. When you’re going for a PhD position, you’re not just signing up for research; you’ll be spending the prime years of your life in a new environment, working closely with a small group of people. The most important of these people will be your advisor. So finding the right advisor for you is probably the most important aspect of your applications.

For this, I recommend the following three ways of finding good people to work with:

1. Look up authors of papers that you’ve read or referenced in your research. There’s a high chance that those people work in your area of interest, and it’ll be a big plus if you have done research in an intersecting area. This is how I found the person who I wanted to work with.

2. Use csrankings. This website essentially collects all the information you’ll need about CS professors and universities. Use this after step 1.

3. Finally, look up the faculty lists of the universities you want to apply to. This takes a lot of time, but is another way of finding potential advisors.

Among the three methods above, I highly recommend the first method. Not only will your research interests align with the person of interest, it’s also a pretty good way to showcase expertise and interest in a specific area, and in some cases can even lead to interactions with your potential advisor even before you submit your PhD applications. It’s also a good idea to reach out to former/current students of the person you want to work with, to get an understanding about their work, lab culture etc.

A note about contacting professors. A common doubt for a PhD applicant is whether he/she should email a professor prior to submission of the PhD application. The answer to this varies; it’s best to look up the homepage of the professor you want to contact, since they usually write about this explicitly.

LORs and planning for the worst

Letters of recommendation (aka LORs) are probably the most important aspect of a PhD application. Choose your recommenders wisely. I would suggest to go with professors who you’ve worked with closely for long periods of time (work which has a duration of upto 1-2 years). It goes without saying that the more successful research you do with your referee, the better the LOR will be.

A common concern for people who haven’t worked with multiple professors is about the choice of the third LOR. Sometimes it’s not possible to find a third LOR writer who works at a university. In such cases, I’d like to point out that most universities recommend two LOR writers to be professors, while the third LOR could also come from someone working in the industry.

Finally, I have a word of caution: ask your LOR writers well in advance (ideally, atleast a month before the application deadline). Also, prepare for the worst: have backup LOR writers if possible. In my case, two LOR writers either refused or didn’t reply to me after saying yes to writing an LOR (due to which my applications were on the verge of being desk-rejected). Send them regular reminders in intervals, and give them plenty of time. Also, do not wait for your recommenders before completing your part of the application. In the worst case, wherein your recommender fails to upload a letter before the deadline, contact the university immediately and ask for a replacement. In some cases, LOR deadlines can be relaxed even upto a month after the application deadline (but it varies across universities).

The interviews

PhD interviews are usually informal and last about 30 minutes to an hour, since your application should be good enough to represent you as a candidate. In any case, some preparation for PhD interviews is always in order. Particularly, know the ins and outs of your research, and if possible, of problems adjacent to your own work. Be prepared to answer technical questions. Sometimes, to make the interviews more structured, professors could also give you some reference papers to read prior to the interview, and may ask you to present the overall ideas of the paper during the interview. Don’t stress too much about the interviews, and treat them as an opportunity to interact with your potential advisor. Ask them questions related to their work, their lab, and their style of advising. This information will go a long way both for you and your potential advisor to figure out whether you two are good for each other.

The final verdict

Most people get their acceptances between late January and mid March. While the process could be stressful, it is important to know that luck is often a big factor in determining the final outcome. No matter how good of an SOP or LORs you have or how good research you’ve done, know that sometimes things could go wrong because of circumstances not in your control. At the end of the day, it is important for a PhD applicant to remember that even applying to a PhD program means that you’re passionate about research and that it’s probably fun to you. Those reasons are good enough for you to keep going despite any temporary shortcomings in your research life :)