The Road to PhD II: My Approach to the Applications

17 min readJun 2, 2023

Introduction

In this article, I talk about the application process and how my application efforts in my first trial in the fall of 2021 and the second trial in 2022 differ. It is important to reiterate my focus on PhD programs in computer science at US schools. My experience might not apply to all audiences' situations if those keywords differ, but some interesting insights should correlate.

How it started and ended

In the first trial, I applied to 10 schools, and I interviewed at 4 of those schools (UW, Rice, Northeastern, and CU Boulder). None of them sent me an offer. It was demoralizing but a necessary learning curve and reorientation for me. Of course, we ended up having an eight-month academic strike shortly after, which meant I wouldn’t have been able to resume if I had gotten any acceptance offers.

In the second trial, I applied again to 10 schools (I swear, it was a coincidence) and interviewed at 7 of them. CMU and Stanford were my first choices due to the impressive number of their faculties working within my field of interest, ML efficiency. CMU was the first to send me an offer. I waited a bit before accepting. It’s a top 1% PhD offer (according to Dr. Schwab, my 2021 PhD application mentor), but I wanted to see what Stanford’s offer would look like and the potential professors I’d be working with. While CMU has the largest ML Efficiency faculty and a unique research culture, and my PI is a professor I feel strongly about, Stanford seems to excel in terms of a warmer climate and proximity to Silicon Valley. I got Stanford’s rejection shortly after (which was a bit surprising since I had a fascinating interview there), and I reached out to the rest of the schools I interviewed at, letting them know I would be moving on with CMU, so they may not bother keeping an acceptance slot for me.

A sample response to my email where I told the rest of the schools I’d be moving on with CMU.

Finding Schools / Labs / Professors

In hindsight, the way I started looking for schools in the first trial was quite comical. I started looking at schools early, around June or July. My search terms were usually “underrated or least competitive schools in the US” since I was unsure how good of a PhD candidate I was. I didn’t know folks around me doing PhD or interested in PhD then. But via Afrisnet, I started getting mentored by Dr. Daniel Schwab. I also met PhD students through several graduate school application support programs who thought I had an attractive profile. Then I published my sign-to-speech work at a NeurIPS workshop and started feeling important. This was when I had enough confidence to go to csrankings.org and look for the top 25, and top 50, CS programs, look through the faculty list, and select schools and professors based on this. Sometime after the application process in March 2022, I was in this Black in AI academic session to reflect on failed graduate applications. There, a professor from Toronto suggested that I look at smaller and less competitive schools next time.

For my second trial, I had been extensively working on my research interests, and I had a long list of researchers whose work inspired or intersected with mine in some regard. I also look at the conferences and workshops to which I’d like to submit my publications and their list of invited speakers. There is a bit of bias in this because most of the researchers whose research publications and talks I see frequently are usually star professors at competitive universities. And subsequently, I applied to primarily competitive schools, even though I didn’t intend to. Of course, a few professors I like are at smaller universities, too, such as Michigan State University and Ohio State University.

Reaching out to Professors

Many debates have been held on whether reaching out to professors before/after you apply is essential. It’s probably negligent, but it may be more critical for international students, especially when your recommenders are not easily recognizable in your target location. Reaching out doesn’t mean one gets a free pass, but a professor can keep an eye on one’s application when it enters the application system. Of course, most top schools and star professors usually detest applicants reaching out to them outside the application system. But a lot of professors encourage students to reach out as well.

In 2021, I reached out to about 25–30 professors. Usually, I’d write about myself, then tell them I am a good fit for their research — sometimes citing 1 or 2 of their works — and they should consider looking at my application or interviewing me. Maybe about 5 or 6 professors responded favorably. Some extra three told me they weren’t taking students. Three out of the four interviews I got were from professors I reached out to, so reaching out was probably worth it.

In 2022, I started my application pretty late. The year was a rough one for me. I wasn’t feeling confident enough to apply for a PhD until Oreva reached out, and we talked, and I also spoke to Annabelle. I started sending out emails to professors in late October (yeah, very late). Usually, I will talk about several of their research projects and how they intersect with my research and future interests. In very few of them, I asked research questions or suggested a future focus on their results related to my interests. The emails are super technical and longish, but I had an unexpectedly high response rate. I wrote about 17 emails in total; 13 professors got back to me, and 11 wanted me to submit applications. Ultimately, I picked six of them that I felt strongly about. The remaining four schools I applied to don’t care about reaching out to professors beforehand, even though I had enough social capital to get introduced to two professors from this category of schools.

Writing GRE and TOEFL/IELTS

I almost wrote the GRE the first time until Dr. Schwab dissuaded me otherwise. Since I was getting profiled for a PhD based on my ability to do research, standardized tests have no footing in that discourse. Usually, US schools don’t require the GRE for PhDs, but some schools might make taking the test necessary for MScs. A good rule of thumb is that a near-perfect GRE score may deflect some attention if your coursework grades are abysmal.

Most likely due to my cynicism, I felt too arrogant to write an English proficiency test. I removed schools from my application list solely on this basis. I consider myself a native English speaker since I have spoken it since birth; it is Nigeria’s sole official language and the sole language of instruction in all of my schooling. Moreover, I have given about two dozen technical talks in English, virtually and physically, in different parts of the world. I feel it is quite unfair for a school to label me as a non-native while accepting a French-speaking Canadian from Quebec as a native English speaker.

Writing my CV

Both attempts at writing my CV cover the following sections; education, research experience, publications, work experience (mainly covering my ML Engineering roles), technical strengths, selected independent coursework, selected talks, honors & awards, and volunteering & community service.

In the second trial, I had a much bulkier research experience section and trimmed down the community services section to a selected few. I used LaTeX code to generate a shorter CV (my final version was three pages) and a less troubling and more aesthetically pleasing sight.

One other important thing I did along this line was create a website. It makes it easy to look me up and provides information about me in a condensed manner. One can easily use GitHub Pages or Jekyll to create a one-page website.

Writing my Statement of Purpose / Personal Statement

My first and second SOPs placed side by side.

This is one of the highly contrasting points between my two application efforts. My first trial’s SOP is a beautifully written thing. It told my story chronologically, and a reviewer could see all the turning points. I started with my interest in CS and AI, my BSc at FUNAAB, and the research project I did as an undergraduate. Then I wrote about how ML Collective has been helpful for my research journey and a research assistantship role that I picked up with a German DFKI researcher. Then finally, my knowledge-sharing efforts via community, my future goals, and why this school is my choice.

This was the only paragraph in the second attempt where I didn’t discuss my research (motivation, experience, or interests). This paragraph was to explain; why my BSc seems so long, why I didn’t perform excellently in some of my coursework, how I obtained my research experience differently from the conventional way, and my other relevant extracurricular experiences, including community services.

In my second attempt, it was more of a research statement than a personal statement. My initial drafts resembled a research article, with citations and all. With subsequent reviews, I found a delicate balance between personalizing it and discussing my research. I split my research interest, ML efficiency, into three categories: data efficiency, efficiency in algorithms & methods, and efficiency in systems & applications, and dedicated a paragraph to each. Aside from the introduction and the closing paragraphs, the only other paragraph was titled “Growing professionally,” and I compressed what could have formed another personal statement on its own into a single paragraph. I barely talked about why I wanted to attend the school, except to mention the list of professors I wanted to work with. In hindsight, my second attempt was too business-like, and it seemed as if I was angry at the world, but I’d probably write it again like this if I were to apply again. You can read why in this wonderful article on how to write a bad statement.

Either style of writing depends on the program one is applying for and the school one is applying to. The first style (the most common) stays perfect for non-CS students more often than not. But AI is an ultra-competitive field and easily has more than 4x the number of graduate school candidates in any other area, and everybody is terribly smart. It makes sense to cut out the fillers and self-serenades and strip to the bones quickly to stand out. How? By talking about the main goal of a PhD: research! ALL the best SOPs I have seen at the most competitive AI PhD programs do this.

Pro-tip: While SOP is solely focused on the research you want to do, it has to be deeply personalized. And all the mentioned research areas must connect!

But if an applicant doesn’t have a clear research interest yet, writing the second way could prove a disaster. It automatically limits the kind of professors one could work with, and you may have schools where nobody works at that intersection. The way I searched for schools and professors mitigates that potential problem. Also, I noticed that the faculty tended to be narrower and less interested in topics outside their field as the school's ranking rose.

I am thankful to all my SOP reviewers at different levels—about 10 of them, for the critical feedback. Big ups to Keegan of CMU, Rajesh of Oregon State, Julius and Sanjit of MIT, Konstantinos of UChicago, Naomi, Nahid, the Black in AI Academics handler, that Stanford PhD student, and that Vector Institute research scientist. A special thanks to Oreva Ahia for the several reviews from her end.

Getting my recommenders

This is probably the most essential part of one’s application, aside from an applicant’s previous research experience. As a general rule of thumb, letters from professors and research scientists count the most. After that, rank letters from lecturers, employers, or postdocs.

I used the same set of recommenders, more or less, both times. I used four recommenders: two research scientists, a senior lecturer from the UK (same rank as an associate professor in the US), and a recommender from my school, FUNAAB. On the first attempt, my supervisor wrote me one. On the second attempt, I used my HoD, the highest-ranking professor in my department. I asked him only because I had helped with some code implementation in one of his large-scale, multi-departmental projects and because he is very familiar with my extracurricular activities due to my collaborations with the department in my capacity as the community lead for the Google Developer Students Club. Naturally, all recommenders are familiar with one’s research capacity and can write enthusiastically about it.

A letter counts a lot more if the admissions committee knows the recommender. Usually, the top researchers in the same subfield know each other. I witnessed two of the professors that interviewed me mention that they are friends with one of my recommenders, and some said that they were familiar with ML Collective beforehand.

Getting Application Fee Waivers

Applying to 10 schools is interesting, but the application fees are definitely not enjoyable, especially if one is from a developing country. Some schools, like Rice, offer free PhD applications. Some schools usually have sessions where they publicize fee waivers. I am immensely thankful for my Black in AI affiliation. Schools like UW, CMU, and MIT offer waivers if you can prove your affiliation with BAI. A few, like Stanford, have a diversity/URM waiver where you have to write about how you are disadvantaged.

Not all schools were covered, and I had to pay for some (3 of the 10) since I feel strongly about all the schools I applied to. In the previous year, Adedayo of MIT paid my excesses, and I am grateful for that.

Submitting Applications

Usually, most US schools have application submission deadlines around December 15. Some are as early as December 1st or 8th, and some are as late as January or February. Filling out the applications repeatedly is always a chore. Since I did not have to change a lot in my SOP, I moved fast this time and submitted all my applications around December 8th.

Interviews

This is straightforward and can be pretty diverse. For instance, I interviewed with two Harvard professors, and both interviews were highly dissimilar. The first one was mostly philosophical: how good ideas are generated and knowing when to persist with or leave a project idea. I enjoy conversations like this, and I remember using Einstein and the Unified Field Theory as examples of the latter topic. The second interview was more practical: what I have worked on, why, and why I am interested in the professor. I had a few, like the one with CMU, where we laughed more than twice. A good number of them were businesslike, and very few were tense. Usually, I was interviewed by a professor of interest, except at Stanford, where the interviewer was only part of the admissions committee.

The bottom line is to be very confident about your research work and be able to defend every single point in your SOP and CV. Do your homework very well. The interviewer is most likely an expert in your field of interest or a similar field, so you cannot bullshit your way through (I tried it in one of my first-time attempts, and it didn’t work).

One of the professors I interviewed with in my first attempt was kind enough to drop a review of my application (I requested it afterward), including stuff I could improve on.

I requested feedback again from one of the schools I interviewed with in my second attempt, and you can see the differences in the review.

In hindsight, I wasn’t convincing enough in my interviews for my first trial. There were too many gaps in my knowledge of machine learning and deep learning. I kept shouting “resource-efficient machine learning,” but I knew little about it. If I had been given anything else, I’d have taken it. I agreed too readily with most conjectures from my interviewers since I knew little (CU Boulder and Northeastern). I seemed a bit confused or overwhelmed on one of the calls (UW and Northeastern). And in one of the calls, I got asked a trap question (“What school will give you an offer that’d make you reject our offer?”) and was too honest. The wiser me deflected a similar question in my second trial.

Selecting Offers

Offers usually roll in around February or March. Hopefully, you'll be able to get multiple offers and juxtapose them. There are a few points to consider when accepting schools’ offers;

The first is the kind of funding. Usually, funding comes in the form of graduate (teaching or research) assistantships or a research fellowship. A teaching assistantship requires you to teach alongside your research, while a research assistantship is tied to your advisor’s funding, so you’d have to work within the field of your advisor’s grant. Either requires 20 hours of work per week. A research fellowship is usually not tied to any obligations and is mainly funded by external bodies like the NSF. My CMU offer letter says I was awarded a school research fellowship. I am not sure what I did to get that.
It is also imperative to understand an advisor’s values, mentoring or research style, and what constitutes a good fit for you. Hands-on vs. hands-off? Confrontational or indirect communication style? Discretion vs. in-group cohesion? Well-being vs. research progress, et cetera.
School visitation or speaking with your potential advisor’s students is also necessary. You’d want to understand what people who interact with your advisor think about them. It’s also not trivial to know more about the program and the living conditions in the school’s city. I had calls with three CMU students around this.

Submitting Transcripts

Usually, it is required to submit one’s (unofficial) transcripts during the application process, and once a school accepts one, it is required for the student’s former institution to send the full, official transcripts to the new institution. CMU allows me to submit my official transcripts at any time before the start of my program. It doesn’t work everywhere that way. A friend had to have her alma mater send the transcripts so the new school could issue her i-20.

It’s not important to submit your full transcripts if you are an undergrad. You can submit everything else aside from your last two semesters in undergrad. Of course, the official, full transcripts have to be received before you resume the program.

Visa application

Usually, the school issues an i-20 sometime in April or May, and then visa applications can commence. There will be a DS-160 form, and you'll have to pay for SEVIS. Then you pay the MRV fee to schedule an F-1 visa interview date. Naturally, lots of people panic about this, but there is actually no reason to do that, especially if you do all this before June or July.

For the interview preparation, there is usually lots of pressure to do lots of mock interviews and get prayerful, but you honestly don’t need to do any of that if you are going for a fully funded graduate program. You may be expected to prove your home ties during your interview. I think it helps to read a few interview transcripts beforehand to understand the kind of questions to expect, but you'll most likely feel understimulated and unstretched when you interview. I read a couple of transcripts via Nairaland and a WhatsApp group, and usually, it is the non-funded and partially-funded students that struggle during the visa interviews. I had a mock interview with my friend, Victor Ojewale (who got into two Ivy League CS PhD programs), but it was mostly banter.

For the interview itself, you don’t have to be there super early. And you don’t have to dress uncomfortably formally. Wearing suits in very humid weather or heels when you'd standing and queueing for a long stretch of time might not be super comfortable. The consular officer probably doesn’t care. I wore jeans and a round-neck T-shirt, and I was fine.

After this, all that should be left are accommodations and booking flights. The school usually provides guides on the former. Connecting and talking to current students too is usually helpful for housing search.

A vote of thanks from a very grateful person

I think it is undeniable that I have a vast support network, and this section does very little justice to appreciating them. As I said earlier, I wasn’t certain I wanted to apply for a PhD again until late October. 2022 was quite draining for me; I most likely wouldn’t have tried applying again if not for the support I received.

I am grateful to all my recommenders across my two attempts (Rosanne Liu, Jason Yosinski, Dr. Babatunde Q. Olorisade, Prof. Olusegun Folorunso, and Dr. A. Abayomi-Alli). They are not only my letter writers but also my mentors. I really hope to make each and every one of you proud.

I am appreciative of all my application materials reviewers, as listed above. I am grateful to ML Collective; this couldn’t have been so beautiful without that open-access research lab. I wish to thank all my research collaborators that helped me achieve clarity in one way or another; Opeyemi Osakuade, Nayan Saxena, Ender Minyard, and Nahid Alam are the most prominent. I am grateful to Annabelle Carell for her support and encouragement throughout my two PhD and AI residency attempts. I thank Naomi Saphra for the referrals and the many (scary) tips about living in Pittsburgh. I appreciate Dr. Daniel Schwab for mentoring me in my first PhD application process and giving me several pointers in grad school decision-making. My thanks to Victor Ojewale for doing this application thing together across the two attempts! :D

Oreva Ahia deserves an entire paragraph of gratitude and praise. My application efforts started with her reaching out to me about assisting me in whatever way I needed. Over the next couple of months, we met between 4 and 5 times, with each call ranging between 2 and 3 hours. Aside from the initial motivation, she conducted several rounds of review on my SOP and CV and had mock interview practices with me. All of this while working on her PhD at UW.

I am thankful to my emotional support group, the ones I tell everything to: my sisters and my foster brother (Grace, Precious, Joy, and Babalola) Kolawole; my roommate, Beloved Temitope; my ASAlytics support group (Busayor Awobade, Ernest Owojori, and Temitayo Oladetoun); Mardiyyah Oduwole; and Abraham Owodunni. I am grateful to Mary Salami for her presence and constant encouragement. I appreciate my close friends in my department, to whom I was a frequent academic liability (Jumobi Joshua, Busayor Awobade, and Dammy Akinosho). A shoutout to Peculiar Abolade and Mary Olawoyin for helping me study for my carryover course.

My first batch of support groups reacted almost immediately after I forwarded the acceptance email to them. :) Oreva’s was the loudest, but those reactions have disappeared on our freemium Slack. :(

I am so grateful to all those who don’t know me personally but whose online contents were exceedingly beneficial. Shaily Bhatt, Paul Liang, and Tim Dettmers come to mind now.

I’d have inadvertently missed some names. Hence, to everybody and every organization that has helped me along this trajectory, including Fab and NACOS, Dr. Bayo Adekambi, and all the community folks at Data Science Nigeria, I am immensely grateful for the assistance. A special shoutout to my former AI Lead, Adeyinka Ogunbajo, who literally dragged me into giving back to our AI communities in late 2019 and who is also resuming grad school this fall in the US.

And to everyone else trying to do some impressive stuff in whatever category, I hope my story has inspired you to pursue your dreams, no matter where you come from or what your background is.

I am excited to begin my journey in this PhD thing. I know that it will be a challenging journey, but I am quite confident in my stocks. I am grateful I am here right now, despite everything I have been through. And I am grateful for all the shoulders I have stood upon in my path to actualizing my dreams.

Appendix

I wanted to write another entirely new article about general lessons and tips for the PhD application process, but I honestly don’t have the bandwidth for that.

A Ph.D. program application is a request for someone to invest around $500,000 and five years of mentorship time on you so that you can produce new knowledge via research publications.

The articles and repositories below contain everything you’d ever need to know about PhD and how to prepare for them.

Other Notes

The surefire way of getting PhD offers? Learn to do research and publish!
Grades don’t matter that much. The most important thing is always your perceived ability to do research in your chosen field.
There’s a lot of randomness.
Social capital matters.