I believe there’s a significant knowledge gap between undergrads and graduate students about the PhD application process, especially for international students applying to programs in the U.S. To help bridge this gap, I’m “open-sourcing” my PhD application experience in as much detail as possible, along with some genuine advice. My goal is to reduce the likelihood of strong candidates getting unexpected results due to this knowledge gap.
What I will cover in this post:
- My PhD Application
- Timeline
- Materials
- School/Faculty selection
- Interviews and offers
- Advice
- How to prepare application materials
- How to prepare for interviews
My PhD Application
Timeline
- Oct 15: Started school/faculty search
- Oct 18: Started working on SoP
- Oct 25: Sorted out the top 10 schools and related professors
- Oct 28: Finalized school choices and potential advisors
- Oct 30: Finished the first draft of SoP
- Nov 11: Finished the second version of SoP
- Nov 15: Reached out to past advisors for recommendation letters
- Nov 22: Sent out all recommendation letter invites
- December: Refined SoP details and submitted all applications
Materials
Here are links to my CV and one of my SoPs: CV, SoP. For all programs I only modified the last section of the SoP, keeping the rest unchanged. My recommendation letters are from three professors I worked closely with. I guess they (or at least their PhD students) knew me well.
School/Faculty selection
I applied to the following schools and corresponding professors of interest (PoIs). All programs are CS PhD programs.
| School | PoI 1 | PoI 2 | PoI 3 |
|---|---|---|---|
| UC Berkeley | Jacob Steinhardt | Anca Dragan | Stuart Russell |
| Stanford | Azalia Mirhoseini | Percy Liang | Tatsunori Hashimoto |
| NYU | He He | Eunsol Choi | Mengye Ren |
| Princeton | Peter Henderson | Sanjeev Arora | Prateek Mittal |
| USC | Jieyu Zhao | Swabha Swayamdipta | / |
| UMich | Lu Wang | Yuekai Sun | / |
| Cornell | Tanya Goyal | Yoav Artzi | / |
Fun fact: I actually planned to apply to UoT but I missed the deadline 😅
Interviews and offers
I got the following interviews. I found professors who didn’t limit interview time to be especially nice! Those conversations went much deeper than other ones. In general, they seem like better potential advisors.
| Date of invitation email | School | Interview time limit |
|---|---|---|
| Jan 5 | Princeton | Unlimited |
| Jan 9 | UMich | 20min |
| Jan 12 | USC | Unlimited |
| Jan 16 | NYU | 15min |
| Jan 16 | NYU | 15min |
| Jan 16 | Berkeley | Unlimited |
| Jan 23 | NYU | 60min |
| Jan 25 | Stanford | 20min |
| Jan 28 | NYU | Unlimited |
Advice
How to prepare application materials
Understand what you want to do. If you are sure that you want to pursue a PhD in the next ≥5 years, you should spend a lot of time to think about 1) what you want to work on and 2) why. Your answers to these questions serve as a foundation for your application: they will be the core of your SoP and the main content of your interviews. They’ll also influence how professors describe you in their rec letters. I think people tend to prefer candidates with strong motivation and specific interest.
Therefore, you should start thinking about these questions early. You probably want to read a lot of related papers to form a good understanding of the field and how your past and future research may fit in. It’s especially helpful to write a short (like 2 pages) but concrete research statements to organize your thoughts. Most programs do not require this, but it can make vague ideas about future research directions much clearer. It also forces you to defend the importance of your work, which in some sense prepares you for interviews. Feel free to share your statement with friends for feedback, even if they’re working on different topics.
Understand importance of different application materials. The most important application material is probably recommendation letters. It’d certainly be good if you can find someone who’s well known in the field to write a letter for you, but I think it is more important that your recommender knows you well. That increases the authenticity and concreteness of the letter, which allows readers to get a more comprehensive understanding of your characteristics and research potential. The second most important thing is probably your publications. You don’t need to have a lot of papers—it should be okay as long as there’s one or more research projects that you proposed, led, or were a primary contributor. Professors will often highlight the quality of your work in their rec letters.
Your SoP comes next, but its importance is muh lower than rec letters and publications. Apart from these, your GPA and courses may also contribute a little bit—some professors may like students who take high level courses—but they are often somewhat negligible compared to rec letters.
How to write a good SoP. Focus your statement on research. There are many ways to write a successful SoP. I prefer starting by clearly stating the motivation and giving a concise 1-2 sentence description of the planned work (and approach, if available). Then, you can talk about your past work and how it led to your current focus. When you describe past work, you should highlight your contribution and implication/impact of the work instead of going through too many method/experiment details.
You’ll likely fill the first 1+ pages with this. Then, you can describe your future work in more details using the remaining space. Lastly, use a few lines to talk about professors you’re interested in. Usually one concise sentence like “Prof. xxx’s work on xxx is aligned with my goal of xxx… because…” or “Prof. xxx’s advice on xxx will be especially helpful because xxx” will be enough.
Start writing your SoP early, and be ready to iterate many times—but keep in mind that you should not stress too much, as it’s not as critical as your letters or publications.
How to prepare for interviews
What an interview looks like. PhD interviews primarily involve sharing past research projects and describing your future research plans. Be prepared to discuss your past projects in detail, and have clear answers for any questions that may come up. Some professors may ask techinical questions (like asking you to come up with methods to solve a specific ML problem, or questions about common ML knowledge), but that’s pretty rare. At the end of the interview, you can ask a few questions.
Common questions. It’s very useful to write down your answers for common questions and practice them before interviews. Examples are:
- Could you talk about your most significant past project?
- Could you briefly talk about all your project and then talk about the most significant one in more details?
- Could you answer this and this questions about one of your project?
- What do you want to do in your PhD?
- What would be your first project?
- What would be your short-term/long-term goal?
- What do you want to do after your PhD?
- What’s your work style?
Some questions you may want to ask are:
- What are some ongoing projects in the lab?
- Work style?
- Compute?
Overall, interviews are pretty chill. If you are familiar with your past work (which you should be), everything will flow smoothly. For your future work, just treat it like a brainstorming with you labmates!
Good luck!