I recently finished my PhD. As the feeling continues to sink in, I’ve been reflecting on it. Five years teaches you a lot, not just about research but about life, and most of what I’ve been thinking about never made it into the thesis.

Each PhD is unique and these are personal opinions, shaped by my own journey and conversations with many people.

Finding research along the way

No one in my family had a background in academia, so I didn’t really have an idea of what a PhD was until later in my undergrad. It changed when I got the chance to do research with great mentors at FOR.ai (now Cohere Labs) and during my MS at KAIST, where I was admitted directly into a lab and fully funded with a stipend to write a thesis.

The realization that you could actually get paid to do what you like was new to me, and I liked it enough to want five more years of it. On top of that, most of the roles I wanted listed a PhD as a prerequisite. I had also spent time during undergrad teaching outside of college hours, which I really enjoyed, giving me an inclination to stay in academia.

Looking back, I can’t emphasize enough how important it is to know why you are committing to a PhD. Your reasons directly impact the kind of work you do, the patience you have, and how you choose your advisor. Here is the set of questions I asked that helped me decide.

The very first project

My first PhD project ended in a dead end. We were probably a bit ahead of our time, but it taught me how to know when to stop a project or pause it for a while. After that, I worked on a healthcare project where I was confused about whether the problem was truly multimodal or not. Kyunghyun (my advisor) asked me a question: “What is even multimodal learning?”

We spent a year or more answering that question. My later work [1, 2] built directly on top of that while also exploring my itch for continual learning through temporal generalization. I don’t think I consciously picked a direction, I just couldn’t put down the question that we started asking. It made me realize that if I’m not genuinely excited about a project and looking forward to working on it over a long period, it’s probably not the right one for me.

Another year of nothing working

For our temporal generalization work, there were long periods when nothing worked. I doubted whether it’d break through. We spent a full year trying to replicate and scale published work only to watch it fail. In hindsight it worked out, but that isn’t how it felt at that time. In those moments, I found it really important to remind myself that I have been here before, that others go through similar phases, and that it is all part of the PhD process. I knew I would get past it.

That period taught me to deeply respect negative results, especially at a time when we have so many publications. The most important lesson I learned was to simply show up. Not every piece of progress will feel monumental; it’s the disciplined, daily progress that adds up until things suddenly click.

I also learned I was worse at multitasking than I thought. Once I kept to one project at a time, it allowed me to focus better and talk about it all the time. I learned to get comfortable with discomfort.

Above all, I can’t overstate the importance of a support network. Most people around you are struggling with some version of the same thing. It only takes someone creating a space to have these conversations, and I wish more of those spaces existed.

If you are stuck or going through a hard time right now, I’m sending strength your way. Trust yourself, and trust me, every PhD student goes through this. You are doing amazing!

I was scared of hard problems when I started. I thought I’d fail, finish with nothing, and waste years of work. Having already failed in the past made it worse. As time progressed, I started trusting myself, my advisors and the process more.

Hard problems also come with their own kind of peace. They feel meaningful, you worry less about being scooped, and stay excited to share what you are working on. It’s not just about pushing a paper past a deadline, but about answering a question you care about that holds value into the future. Over time, you build your unique research taste, which defines your work. I believe that’s when you know you can graduate.

NYU and New York

Having never been to the US before starting my PhD, I had no real sense of the physical locations of the schools I applied to. I got incredibly lucky with NYU. The city and the university offered so much that I grew in so many aspects of life, whether that was hobbies, food, coffee, music, theater, comedy, or the list just goes on and on :)

I needed that more than I expected to. Work-life balance was especially challenging during the harder phases. When you are working on something you care deeply about, striking a balance is tricky, but essential. I’ve tried to remain mindful of my mental health throughout my PhD, and I always attempt to maintain active hobbies to pull me out of work.

Why not just go straight to industry?

I value, understand, and appreciate the work done in industry. Industrial labs do fundamental research, and many of them train people well through residencies, internships, and mentorship. Some of the questions I care about are being asked there as well.

But I don’t know many places that would have paid me to spend a year on the definition of multimodal learning. Academia is ideally supposed to be not for-profit. It has constraints, but the freedom allowed me to work on hard, long-term problems without worrying about turning a profit. I found peace in it.

The other difference is structural. Training students is what a university is for, so it survives shifting priorities. During my PhD, I learned to present, to communicate progress, to be patient, and to defend what I did. Industry can teach these too, but academia has a system whose sole purpose is to build those skills in you until you graduate.

Teaching in the realm of AI

As a Teaching Assistant and mentor to students during my PhD, I have constantly thought about how education must evolve alongside AI. Completely stopping the use of AI isn’t the solution, nor is blindly adopting it. There has to be a balance, and we need to lead the way in finding that middle ground while asking students to trust us in the process. The hard part is that we are working it out at the same time as we are asking them to trust us on it.

Would I do it again?

Doing a PhD is a privilege, and I am grateful for it. The stipend may be modest, it’s not easy, and it contains lots of days when you have to work continuously and navigate hard times. But if you are fortunate, it gives you five or more years to work on what you truly care about. It shapes you not only as a researcher, but as a person. Learning not to fear hard problems and building the perseverance to push past them is a skill I cherish.

The curiosity you develop allows you to view the world differently. At times, I feel like a kid again, getting the chance to answer questions just because I’m curious. Even in daily life, I find myself asking questions no one else might care about, but I find beauty in that. Looking back, I wouldn’t trade this experience for anything.

Looking back

A PhD is a marathon, not a sprint, so please reach out to your peers and share your experiences whenever you can. You are not alone in this, and a PhD needs to be an experience you cherish, not dread. Many of us live far away from our families and it can get lonely, so lean into your network, uplift each other, take breaks, and feel proud of the work you do.