Learning to become a scientist might have to do more with the way of thinking than an occupation. When you truly embrace this identity, you learn to think independently. Each paper becomes a snapshot of your thinking at a particular moment. You start operating in a slightly different world, possibly making it difficult to relate to others who don't share this way of seeing things. This can be both liberating and overwhelming. I wanted to take some time to reflect on why I chose to walk this path in the first place.

I published my first paper in high school, back in 2019. I was in the school physics team named BaCoN, after Francis Bacon. The name was fitting for what we believed in, Bacon was critical of Aristotle's approach to natural philosophy, particularly the reliance on deductive reasoning from first principles, which was overly theoretical and insufficiently grounded in observation and experimentation. I've moved away from physics now, but the empirical spirit still guides the way I think.

I realize that learning to do research taught me a way of thinking. There's something deeply satisfying about scientific method; forming hypotheses, testing them through observation, and sharing your discoveries with others. Especially nowadays with everyone's attention span getting shorter due to social media, having someone read your work carefully and think deeply about the same problem together feels precious. Then, there's this aspect of building on top of humanity's knowledge. Writing and publishing become more than just a task, it becomes part of a long tradition of knowledge discovery.

My research career only took off when I started working for an NLP startup in 2020. I took a leave from college. I was also away from my parents because I decided to live by myself right after high school. I was first hired to do some simple frontend development tasks and was randomly tasked to research a system that can measure the readability of a text, mostly because everyone was really busy and I had some physics research experience from high school. I ended up proposing (with presentation slides) that we should submit to an international conference and that our goal should be EMNLP. The team was skeptical at first, but we found a way to make it work together.

Well, now that I think of it, it was ambitious, fresh out of high school, proposing to submit to a top conference when NLP was already incredibly hyped. Nevertheless, we ended up developing a SOTA system and got published. I ended up leading most research projects in the small NLP startup, which helped a lot in building my career. I just really enjoyed experimenting and writing it up, contributing to more than 10 papers and patents in my two years there. I've gotten several questions about how an undergrad got so deeply into research, and this is how it happened.

One thing that I regret is that I should have sought mentorship much earlier in my career. Reading the literature thoroughly and finding people who can mentor you can help you waste less time. Ultimately what matters is how close you get to the truth, not just the number of papers. But truth is like a proof; you can follow someone else's steps, but understanding why each step works requires your own insight. The best mentors know this. They guide you to discoveries rather than simply handing them over.

Research, in essence, is the art of producing new and meaningful knowledge for a specific audience. A single piece of research can enlighten a few readers and a sequence of well-timed insights can redirect an entire field. Of course, you can't always predict the impact. Therefore, having some sense of mission, even a rough one, helps guide your choices about what to work on. Because once you publish, it's permanent; your thinking becomes part of the record that future researchers will build on. And that's worth embracing, even if the path sometimes feels overwhelming.