Chenyu Zhang | 2024
[Edit 1]: I am an international student who applied to US PhD programs in machine learning and applied math. Since I have no experience outside this context, this post may be most relevant to CS and applied math related programs, though I hope it’s still helpful to a broader audience. Please note that this post reflects my own experiences and perspectives—consider it as one piece of information in an online community.
Sorry for being late to the party. This spring, I wrapped up my PhD application cycle, and now, as I sit in my office at MIT LIDS&IDSS, I feel incredibly fortunate. The application journey taught me more than I ever expected—things I hadn't known or wouldn't have believed before going through it myself. In this post, I want to share some insights and lessons learned, especially for those of you who might share a similar background. Please note that these reflections are from my own experience and perspective, and I'll touch on a few observations that aren't widely discussed or universally accepted.
[TOC]
My background is decent but far from the top tier. While I'm confident that I was qualified for the programs I applied to, I still feel incredibly lucky to have received offers in today's fiercely competitive environment in EECS and Applied Math.
Looking at my background, it's clear that I have a relatively weak academic record: My undergraduate GPA is on the lower side, and my Master's program is "awkward"[1]. My research experience is decent, but when I applied, none of my papers had been accepted yet (they were under review with preprints available). My CV at the time of application is available here.
By now, you might have a sense of my chances of getting into these programs. To be honest, while I believe I'm qualified, I know my background alone wouldn't have been enough in today's fiercely competitive landscape. An oversimplified explanation is that I got these offers because of the connections, of my advisors.
I had five strong recommendations. By strong, I don't mean that they were extremely positive or came from well-known professors. I mean that my advisors were willing to leverage their connections to help me secure these spots:
Nearly all the interviews and offers I received were from programs or PoIs that my advisors had connections with. Conversely, for some less competitive programs where my advisors had no connections, I was rejected without an interview. In one CS program interview, the PoI told me he was sifting through around 6,000 applications. Whether or not that was an exaggeration, it's clear that filtering by recommender names is easier than scrutinizing every CV.
Before I applied, I couldn't believe that connections could have such a decisive impact. But the stats and my experience speak for themselves. If you were a busy professor, how would you respond to a cold email from a student you don't know—assuming you even had the time to read it? Would you sift through every application if trusted colleagues had already recommended their top students?
Understanding the weight of connections (of your advisors) is critical. Then other questions arise: How do you find well-connected advisors? How do you make your advisors leverage their connections for you?
One of the most challenging aspects of the application process is gathering and interpreting all the information to make informed decisions. When I was applying for my Master's program, I worked with an agency. They made things easier by handling program research and deadlines for me. However, agencies can be biased, focusing more on getting you accepted rather than finding the best fit for you.
When I sought help from my old agency for my PhD applications, they simply advised me to consult my academic advisors. Your professors, who know you well and understand the academic landscape, are invaluable sources of information. Additionally, asking your professors "what programs should I apply to" is actually asking "which programs do they have connections with?" For example, while nearly all senior graduate students recommended me to apply to Berkeley IEOR as it's relatively easier, my advisor said "If I were you, I would apply to EECS." Following that advice, I successfully gained admission.
So advice from senior students can also be biased, as they have only gone through the process once and may favor their own experiences. The pros are that their insights tend to be more recent, candid, and drawn from their own journeys and resources. When seeking suggestions from senior students, you might encounter strong opinions like "don't pursue a PhD" or "you'll never get into this program." It's important to critically assess which advice is objective and which is influenced by personal biases.
I saw too many students spending too much time crafting their statements, struggling to make them perfect, eventually hurting the other parts of their applications. Just ask your professors and hope they are honest. My advisors told me they would only look at CV + letters. Based on my experience and other sources, here are some key points:
A rule of thumb is to always consider yourself as the (stupidly busy) PoI reading the statement. Would you invest time in a two-page statement if you had 6,000 applications to review? Would you be interested in how an applicant discovered their passion for research in childhood? Or would you be more interested in a writing sample that demonstrates the applicant's ability to produce future papers and proposals?
Unless you're absolutely certain about your research direction and less concerned about finding the right fit with specific programs or PoIs, it's wise to be somewhat vague about your research plan. During one interview, I mentioned that my research interests closely aligned with the PoI's publications, hoping to flatter them. They responded, "Oh, I'm not interested in those topics anymore. I wrote that paper because of my expertise, not because of my current interests."
There's often a gap: you're seeking advisors whose expertise matches your interests, while good professors are continually exploring new topics and prefer to be guided by their students' fresh perspectives. Therefore, the key is to emphasize your research ability rather than a rigid research plan. Expertise and a specific plan are byproducts of your research maturity; having a concrete plan can sometimes limit your opportunities.
Even if no one else reads it, your statement is crucial for you because it helps organize your goals and preparation for graduate school. Since deciding to apply for a PhD during my undergraduate years, I maintained a draft of my SoP, updating its outline each time I participated in a professional event. Before starting any project, I would ask myself, "How does this project fit into my SoP? What experiences and outcomes do I need from this project to grow my SoP story?"
Maintaining a running SoP outline offers several benefits:
The key idea is a familiar one: view your application as a cohesive whole. While your statement might not be scrutinized by everyone, its story should be well-organized and logically connected to support your overall application.
Again, what matters most is your research ability. Your research ability is like a hidden model, emitting signals like publications, grades, and recommendation letters. The emission kernel for publications varies by field.
For instance, publishing a paper in math or theoretical CS is exceptionally challenging, and achieving this in a prestigious journal is even more difficult. Conversely, in application-oriented fields like empirical AI/ML, having multiple publications in top conferences is often a common requirement (for top programs).
I know several students who were admitted to MIT without any publications. They either
Therefore, if obtaining publications is a challenge, focus on demonstrating your research ability in other ways to the PoIs; although this is in no way easier.
Beyond showcasing your research ability, publications can be a powerful tool for building connections. Professors are notified when their work is cited in your publications. This is a far better chance for your name to be remembered than sending them tons of emails. For example, instead of cold emailing a PoI asking them about the application process, you could say, "Hi, our recent paper published in a prestigious venue was directly inspired by your work!" This approach not only grabs their attention but also highlights the relevance of your research.
This also leads to two key takeaways:
When selecting programs and PoIs, both before applying and after receiving offers, there are many tricks and dimensions to consider:
Many applicants focus solely on popular programs and well-known professors, overlooking these critical dimensions. This narrow focus can lead to lower success rates. Consulting your advisors and senior students about these factors can provide valuable insights. You will likely get surprisingly different answers from them though, even from the same person before and after you've received offers. For example, senior PhD students might highlight the prestige of working with a famous professor, while junior students often express frustration with their hands-off approach.
Once you have offers, reach out to senior students in the research groups you're considering. Usually they might give you general, positive feedback like "I'm happy here, the group is nice," as no one would directly call their advisor a d*ck. It's important to ask targeted questions directly and wisely. For instance, "How do your advisor's connections help you secure internships?" This approach will give you a more realistic picture of what to expect.
Thank you for scrolling through this lengthy post. Feel free to share in the comments any questions, suggestions, or disagreements. If you'd like to connect or have any further questions, don't hesitate to reach out to me. Good luck with your applications!
If my reflections have been helpful to you, feel free to buy me a coffee. Cheers 🍻!
Quoted from one of my advisors. This program is industry-oriented with a relatively low academic reputation.↩︎