The Road to PhD I: My Background and Motivations
Introduction
In the first installment of my articles on how I “found” myself in a PhD program at one of the world’s best computer science (CS) research institutions, I talk about my background and the experiences that shaped me and my trajectory into this current state.
Here is a disclaimer: this article is more about my background than the PhD itself. But I feel it’s necessary to understand where I am coming from to avoid falling into the fallacy of hasty generalization.
Cumulative advantage is the hypothesis that the more privilege you had in life, the more likely you had the resources (money, location, educated parents, mentoring, good peers, free time, extracurricular activities, extensive social network) to do well (rapid development, good grades) and this gives you more resources (better schools, better jobs, better connections) to do even better (promotions, tenure, grants) which yields even more resources (even more extensive social network, collaborations, funding) to do even better (Nobel prize, unicorn startups).
I struggled to find the proper way to synthesize this in writing. Besides, several people have expressed interest in knowing more about my trajectory, so I pondered writing this part as an FAQ. A shoutout here to Chidinma Nwatu for suggesting what a comprehensive FAQ about my background and motivations should look like.
- Are your parents professors? How did your background influence your career decisions?
No, I am a first-generation college graduate. Neither of my parents went to college; both of them were affected by one of the ills of the legendary, overpopulated African polygamous families, but that’s another story entirely.
And we were poor, too. One of my three sisters, Precious, wrote a bit about it where she describes her journey from dropping out of OAU to resuming at Shopify and Carleton in Canada. Still, I probably experienced destitution more vividly since I am the eldest. I can still remember (plenty of) days we struggled to eat consistently and days we were sent home from our elementary school for non-payment of school fees. I remember street hawking very briefly when I was 7 or 8. There was a lot of hardship, but everyone goes through their own unique hardship. We slowly became middle-class at some point, but that didn’t happen until the first three children had matured (and we matured pretty fast). But the hand-to-mouth experience did help us in several ways, like the one below.
My siblings and I were very self-sufficient from a very early age; we all had to. And being the eldest and the only male, I started the (consistent) self-sufficiency journey way earlier. I was ten years old and had just resumed my high school education at Abeokuta Grammar School. Around my second term in high school, I went with my dad to visit a friend of his (I still call him Oga today) who owns a block industry somewhere opposite my school. I took a fancy to the workers making the bricks and joined them playfully. That was how I signed up for years of employment in the block industry until I finished high school. I worked part-time during school days and full-time on Saturdays and holidays. I realized that I could buy myself all my favorite bread and beans and the school accessories that I wanted, and by my 3rd year in high school, I was basically taking care of all my bills except for housing and the food at home whenever my mealtime met me at home. It was a harsh working environment, and it took me a while to adapt and “toughen” up, which meant I was the easiest runt of the litter to pick on until I adapted. We regularly exchanged curses and fistfights; occasionally, we had shovel fights when tensions were high. I remember engaging in a shovel fight myself once. Also, some of the boys periodically indulged in certain “recreational” activities to help them stay high and unwavering in their work efforts. Still, I think that was the beginning of my work ethic and grit journey.
The second childhood experience was my reading culture. I was a bibliophile (I wrote a tweet about it). I believe it started when I found Dad’s chest of books, which were an eclectic mix. The books range from ideology-, fairness-, and nationalism-focused books by African authors in the 15th–17th centuries to Western (primarily British) authors (whose writings tend to be less intense and more laidback) covering stories and cultures from Europe and America to the Middle East, Asia, and several parts of Africa — in short, all of the colonial British empire at its greatest. One of my favorite books in the chest was the autobiography of Robin Knox-Johnston. His grit and ambition, which weren’t for the money, were a wonder, at that time, to a kid like me who was always in some state of lack, and in the countless times I reread the book, I’d imagined myself going on that same voyage that Knox-Johsnton went on and winning that level of prestige.
At some point, I had gone through my collections of books multiple times, and then I started borrowing books from people. I didn’t have many friends because I was a 99% introvert. Also, buying books was a luxury back then. So when I couldn’t borrow them, I stole them at times (I had solid moral scruples, I promise, but I just had to read). One time, a bookstore near my workplace had its roof destroyed during a downpour, and I was fortunate enough to inherit all the partially damaged books. I read many of the world’s classics (including the entire collection of Shakespeare and most of George Orwell’s popular books) and so much more across a wide range of genres and topics. I’d usually read my non-school books deep into the night using oil lamps, then later with battery lamps when they became a thing. My eyes’ extra sensitivity to white light might have something to do with this. Reading with electricity-generated light was very rare—not that our neighborhood didn’t have an electricity connection; it was just epileptic. I was an obsessive reader, and I read in the most unorthodox of places—including at the back of speeding brick delivery trucks, with one hand holding the side rails and the other holding a book. I started reading online and surfing Wikipedia when I got my first mobile phone. The reading obsession only abated when I transferred it to computer science. Needless to say, all these readings shape my intrinsic values and how I view the world.
When I started computer science, it wasn’t just about my work ethic. What we work on and who we work with are pretty essential. I obtained the exposure and clarity to make decisions through my compulsive desire to read about the knowledge I desired. I can posit that both contribute well to my intensity and career trajectory.
- How did you decide on computer science (CS)?
You can already deduce that I wasn’t born with a computer in the placenta, nor did I start coding when I was 4. In fact, I didn’t touch or use a computer until after my secondary school education. My school had a computer lab, but I guess it wasn’t available to students because of potential maintenance issues and costs (I assume that’s why we never opened our book-filled library too). We’d usually go to the lab to watch PowerPoint presentations without touching anything, and we’d usually draw and label parts of computers and Microsoft Office packages in exam scripts. We drew many logic tables and wrote lots of BASIC, but we touched no actual computers. Realizing that I could pursue CS in college happened rather fortuitously.
I was done with my high school education and had switched from working in the block industry to working at a steel company (which required more expertise, but the pay was higher, and I had to start saving for college). I was considering studying civil or mechanical engineering without absolute conviction. I was book-smart to a reasonable extent, and I had rigorous technical drawing experience from the last three years of high school (the TD students from Abeokuta Grammar School will remember Baba TD, notorious for his strictness and superior cane-handling skills, fondly). Hence, the standard career convention in Nigeria postulates that engineering is the perfect fit for my profile.
Almost a year after high school, there was a Sallah festival (Eid al-Adha), and my siblings and I decided to visit one of our Muslim aunts. Coincidentally, my uncle-in-law transitioned from his microbiology degree to software development. We chatted about my career paths, and he suggested studying computer science. I agreed and decided to rain-check at his computer training institute (one of the few standard ones in Abeokuta in 2015). This development was fortuitous because I wasn’t close to my extended family and hadn’t seen them in years. Visiting that particular day was also rare because I’d usually be at work, even on holidays. I am glad I did on this particular one.
I went to the computer training institute and realized I liked this computer thing, so I bought myself my first computer. Then I took an indefinite leave from work and started my digital literacy journey. I decided to write JAMB and study CS in higher education. My first trial was only able to land YABATECH, which I refused to take up, and I got into FUNAAB for CS in my second trial. All in all, I spent almost four years between finishing high school and resuming my bachelor’s degree program at FUNAAB.
I filled that period until the end of my first year in university with lots of reading and studying what CS could offer me. Wikipedia, Quora, and subreddits on CS and programming were favorites, and they provided me with lots of clarity (and clarity is the significant upside for me). I always knew what I needed to do to become the best in the world at what I do (speak of white knights in pursuit of glory). For example, in my first semester of 100L, I already knew that interning with Google was (and still is) one of the most prestigious hats a computer science undergrad could wear. I wanted it and even pushed for it. By the beginning of 300L, I was already confident I’d be pursuing a Ph.D.
- How did you decide on artificial intelligence/machine learning (AL / ML)?
Toward the end of my time at the training institute (I left and went back to my iron-bending work when I started running out of cash and still needed to save for college), I started front-end web development. Frankly speaking, I didn’t like it a bit then; I'm not sure why now. So when my school chapter of the National Association of Computing Students (NACOS) started a training program for junior students, I didn’t bother joining the frontend and backend (by association) tracks. Through my readings, I knew that mobile development is almost like front-end web development, so I also didn’t join that. I was big on cybersecurity, but it wasn’t one of the offered tracks, so I joined the Python programming track instead. Most junior students didn’t pay much attention to the extra coding classes due to a more significant focus on their coursework, but I was fortunate to do the opposite. Soon, I had our facilitator, Fab (who was simultaneously the most community-oriented developer in FUNAAB at the time), almost all to myself (even after the classes broke down due to a lack of commitment from both the facilitators and attendees), and I am grateful for the opportunity.
I became strong with Python and decided to skill up further. Data science was the new kid on the block then (at least in my neighborhood), but it sounded exotic, so I decided to try it out and loved it. Since data science is more than just the machine learning part and is also big on data analytics—which I wasn’t too keen about, considering my strong programming background — I slowly gravitated toward ML and AI.
But ML didn’t have many employment opportunities to grow as a Nigerian, and that is where most of my data science friends crossed over to software development. I almost pivoted myself. However, Fab encouraged me to look beyond the Nigerian market and focus on building myself until I was ready for the international scene. I leveraged more on hackathons (I wrote about how I qualified for an AI Bootcamp in 2019, won a gold medal at another 2019 hackathon, and lost at the final round of a Jan. 2020 hackathon), took up several freelancing projects that caught my fancy, did 4 or 5 internships, most of them unpaid, and ventured into open-source development to experience a high-quality work-like culture. I even tried twice to participate in GSoC but was unsuccessful both times. The second GSoC attempt was quite painful because I had planned the second half of that year around doing GSoC and leveraging the experience. Still, those experiences were career-defining for me.
- How or when did you decide to pursue a PhD despite being an undergraduate? Which of your experiences motivated you to go for a PhD?
I had known from a young age that I wanted to be a scholar. Unfortunately, it wasn’t for any very well-thought-through considerations. Naturally, being the bibliomaniac that I was, my folks started calling me “Wole Soyinka” (after the first black African to win a Nobel Prize in Literature). Wole Soyinka is from my city and went to the same high school I attended before he relocated to Ibadan. He lives reclusively in a forest just a couple of miles away from our neighborhood, so it was straightforward to call me that. Again, naturally, I consumed a good number of his works, but the similarities ended when I joined the science department in senior secondary school instead of the arts. Of course, I kept wondering for the longest time if becoming a science student was the right decision for me, considering my reading flair, and that feeling didn’t abate until I got hooked on CS.
When I started seriously pondering data science, I realized that it required more advanced learning than traditional software engineering (SWE) fields. Most (international) data science roles then required PhDs, and many Kaggle Grandmasters were PhD holders. Then I attended Data Science Nigeria’s AI Bootcamp in late 2019, and a significant number of the world-class instructors had specialized degrees, which made an enormous impression on me. My decisive factor was reading Kaggle Grandmaster Vladimir Iglovikov’s interview with Sanyam Bhutani sometime in 2020. In fact, I wanted to build my research experience by leveraging Kaggle until I mistakenly found ML Collective, but more on that in my next article.
Remember how I read a lot? Hence, when I started reading up on the possibilities of a PhD, I realized that US schools (and a few UK and Canadian schools) usually offer PhD positions without an MSc prerequisite if the student can demonstrate substantial research potential. However, it is typically a five-year program instead of the usual four-year PhD in Europe. Then I realized that research-oriented American undergrads apply for PhD in their penultimate years of college. I was just starting my 3rd academic year then, and I decided to apply for a PhD if I could secure quality research experience before I finished my 3rd year. By the middle of 2021’s second quarter, I had started working on my sign-to-speech work. By the final quarter, I had expanded the work into a full-scale research project and had published it at a NeurIPS workshop (I wrote in detail about my research evolution in my 2021 review article). Of course, I applied for a PhD late that year and didn’t get in, but that is another story.
- Are you a first-class student?
I am not. I finished with a 4.1 out of a 5.0 CGPA. I think even with a CGPA less than that, for example, 3.0 out of 5.0, you should be able to get a fully funded PhD program. Grades are nice to have, but the ability to do research is the most critical prerequisite.
- Did coming from FUNAAB, an agriculture-specialized school in Nigeria, affect people’s perception of you in your application?
If you mean in comparison with the top and well-known schools in Nigeria like UI and UNILAG or with a technology-specialized school like FUTA, not at all. Nobody cared about that (except me in the initial days of putting my application together). Of course, this might not hold water for other fields aside from CS; I have noticed that UI seems to have the most significant number of Nigerian schools’ alumni studying in the Western world.
Compared to an average research-inclined American-taught student, it definitely matters. Being an international student from one of the top European schools, the C9 League in China, or the IITs in India might ease things a bit. But if one is not from either category, one probably has more to prove.
- You seem to have had lots of extracurricular experiences. How did you balance that with your coursework?
TLDR: There was no balance.
There was a time I wanted to work toward bagging a first-class because I like to think I am first-class material, and I genuinely love CS. But in the end, my desire to pursue my interests, gain technical depth, and independently seek practical experience was more potent than my attempts to score excellent grades. By the end of my 200L days, I had focused disproportionately more on my interests and never recovered. The most considerable influence was the prolonged academic inactivity due to COVID lockdowns and the unions’ strikes. I gained admission in late 2017, and I was supposed to finish the degree in the 2020–2021 session. But I didn’t resume until mid-2018 and didn’t finish until 2023. Naturally, I filled my time with an excessive amount of independent work.
Implausibly, I think the technical depth helped boost my intuitive understanding of the courses we took later, and I could afford to spend less time attending classes and cramming for exams. Giving instances;
- By the time we did the Structured Programming course in 300L, I was proficient to a reasonable extent with Python and had facilitated dozens of Python tutorials.
- By the time we did the AI and data-related courses in 400L, I had taught hundreds of AI enthusiasts physically in Abeokuta and virtually mentored several more across Nigeria and West Africa. I was even the Nigeria Computer Society’s AI Champion that year.
- When we did network programming and built dummy servers with Socket, I had already done a research project on federated learning (a privacy-preserving focus for client-server architectures) and written a research workshop paper.
- Around the time we studied courses titled Entrepreneur for CS, Net-centric, and Formal Methods & Software Development in school, my friends and I had designed a fully cloud-based ML/SWE architecture, written a grant proposal (detailing our value proposition, engineering & pipeline design, engineering decisions & tradeoffs, milestones, and an operational guide) for it, and been awarded a grant in six-figure USDs.
If the Nigerian lecturers’ marking system were not a bit rigid and slightly random at times, my CGPA might not have looked like I was struggling to stay a 4-pointer. But I probably deserve my grades. Focusing on my interests is a calculated risk, leading to a downward spiral with my coursework that could have gone out of control. From 300L until I finished, my interest in most of my coursework diminished exponentially. I barely attended most classes, did not complete assignments, and developed a reputation for appearing late at everything, including the seminar presentations and exams. Some of our faculty respected me for being ahead of the curve, and a few lecturers resented me for not taking their courses too seriously. A few were quick to call me out on my “pomposity and unseriousness.”
Moreover, I feared that I might fail some of my courses and have to spend an extra year, but that fear was minuscule. Nevertheless, focusing on my interests was one of my better decisions. But I won’t deny that my technical focus is closely tied to my education, which was a boon. While CS intersects with virtually every field, this wouldn’t have been so straightforward if I were a mechanical engineering or linguistics student.
- Did your community experiences play a part in this journey?
I initially forced myself to engage in community participation and public speaking to eliminate my deep introversion and social awkwardness. I realized I’d need social skills to advance my trajectory. But it became more than that—I am a product of community development, from Fab via NACOS to Data Science Nigeria’s (DSN) vision to unearth 1 million AI talents in Nigeria, to the extent that my research trajectory was powered by the wings of ML Collective. Most of my closest friends are from my communities as well. Community is tied to my entire being. And research itself is a community-focused ideology.
Regarding the PhD application process, my soft skills certainly helped in my social interactions, leading to a PhD offer. However, I am unsure if any of my employers or schools care strongly that I was a DSN Ambassador, a Community Lead for the Google Developer Students Club, or that I care about female empowerment, as seen in my mentoring at She Code Africa. Regarding the weight it added to my PhD application, I would chuck it directly under my grades in my hierarchical list.
- Looking at your trajectory, you could have easily focused on ML engineering instead of research. What were the deciding factors?
This is due more to luck and available opportunities than overly careful planning. This peregrination might best be modeled by my three efforts to work at Google (remember I mentioned earlier I wanted Google because interning there is a prestigious hat). All three tries were with referrals (although social capital and job referrals are outside the scope of this article). I detailed the last two tries in my 2022 review article.
My first attempt was for the (ML-focused) software engineering internship at Google in 2020. I was chatting casually with Jonathan, a Google accountant who mentored one of my sisters, Grace, at YaaW, and my ambition to intern at Google accidentally slid in. :) Jonathan was kind enough to offer me a referral. I felt ecstatic! He reviewed my CV, and I got two other Googlers on LinkedIn to review my CV. I submitted it, and I never heard back from Google (the same CV got into the interview stage at Microsoft a few months later, but I couldn’t make it to the final interview stage). I felt stupid!
Then, in late 2021, Abdoulaye, AI Program Manager at Google Africa, generously referred me for the AI research residency program at Google Accra (he is a fine man; by the time we met in Tunis in late 2022, he had forgotten about it). This role requires a mix of software engineering and research skills. I got into the interviews this time but didn’t make the final cut.
The third one was the student researcher (BSc) role at Google Brain, San Francisco. Rosanne, my mentor and senior research scientist at Brain submitted an inside application for me in 2022. I passed through the (purely research) interviews this time, but Google announced a hiring freeze before I could settle on a team to join and get my formal offer letter.
It is apparent how my work goals morphed from engineering to research roles over time, which is primarily serendipitous. ML engineering roles are hard to land, mainly due to my location. I described in my 2021 review article my experience trying to find a quality role for my six months of SIWES training. But open research transcends location (thanks to communities like ML Collective, Masakhane, and C4AI), and that was what did it for me. On my sign-to-speech project, all four of us were in different locations and countries. I was in Abeokuta (you’d probably notice that I rarely leave my home city by now), Opeyemi was in Lagos, Nayan was in Toronto, and Dr. Qasim was in Cardiff. Focusing solely on research also meant I had to work, driven exclusively by my research motivations, for a long time without a company’s affiliation to augment my living expenses. And wrapping this up, I am giving a shoutout to my sister, Grace, for being my biggest breadwinner during the earliest part of this phase.
