Master of Science in Computer Science

Late-night study sessions, breakthroughs, breakdowns, and way too much coffee. With my degree in hand, I was exhausted but ready. Time to go build something.

My interest in deep learning didn't end with Yann LeCun's course. I ended up interning at VTracker, doing semi-supervised learning for perception models in maritime object detection and tracking. Built an entire framework for them to use on their ships.

This was our thesis for a game development course. The niche was overlooked: creating aesthetically pleasing chess puzzles. Sure, there are AI solutions that beat chess way better than humans, but almost no work had been done on puzzle generation. We built an algorithm that, given any position and a set of heuristics, could create beautiful puzzles. We backed it up with data too. We ran a survey, asked real chess players on chess.com to rate puzzles, and our generated puzzles scored way better than the previous SOTA. They were close to Lichess's official puzzle list, which is a big deal because those are human-made puzzles that are meant to be puzzles. Ours were being generated in seconds.

I got to learn from Yann LeCun, the star professor at NYU. He introduced me to transformers and energy-based models, which took my CNN expertise way further. That eventually got me into 3D perception models and shot up my interest in autonomous vehicles. I even ended up building a prototype for simulating a vehicle.

One of the most interesting courses I took wasn't your traditional textbook course. Heuristic Problem Solving was hands-on, literally. Every week we'd get a problem statement, write a solution, and the next week compete against someone else from the class. The winner gets KitKats. I won most of those challenges, and my team would always end up near the top. But there's this one story I just can't forget. We had a gambling problem: act as a casino trying to make money, and also act as a user trying to beat the house. We were writing these complex algorithms with probabilities and heuristics. The team that won? They just chose not to gamble at all. Their entire solution was return 0. Sometimes the simplest answer is the best one.

For my internship I had a lot of options: a banking project, an ML startup doing OCR, big data stuff. Out of all of them, I chose a crypto startup (I still have no idea why). I'd never really understood blockchain, so maybe I was just trying to see what the hype was about. We were building NFTs with real physical value, taking collectible assets like cars and Pokémon cards and making them tradable. That's where I picked up React and frontend for the first time.

After Cassie, I continued pursuing my Master's in CS at NYU. I was super stoked to learn from the best faculty, and the shift in curriculum, the focus on assignments and research, was definitely a challenge to adjust to. But it was something I was willing to push myself for. The coursework was rigorous, and I got pushed to my limits.