It was 2 AM on a Tuesday in my second year of college. I was procrastinating on an assignment — and I accidentally stumbled onto the video that rewired my entire future.

The Video

I was watching random YouTube videos to avoid studying. One recommendation after another, until a thumbnail caught my eye: "How Netflix Knows What You Want to Watch." I clicked it expecting a short explanation. 45 minutes later, I was still watching — learning about collaborative filtering, matrix factorization, and how recommendation systems work at scale.

I remember thinking: this is math I've seen before. This is code I could write. And it's behind the feature billions of people use every day?

The Rabbit Hole

That video led to another, then to a GitHub repo, then to a research paper I barely understood, then to Kaggle, and somehow at 4 AM I was reading about neural networks and asking myself why nobody had told me this existed before.

"I realized AI wasn't magic — it was pattern recognition. And patterns are just math. Math I already knew."

I downloaded Python the next morning, enrolled in Andrew Ng's ML course that same week, and built my first small model — a house price predictor — within a month. It wasn't good. It barely worked. But the feeling of a machine learning something from data was unlike anything I'd experienced in my CS curriculum.

What Changed After That Night

Everything, slowly. I started approaching every college project through an ML lens. When we had a database project, I asked: "what if we used this data to make predictions?" When we had a networking assignment, I asked: "what if anomaly detection was ML-based?"

By third year, I was deep into NLP, working with large language models before they were cool, and eventually started building the projects you see on this portfolio — the Bayesian medical AI, the RAG system on AWS, the AI resume parser that won our college hackathon.

All of it traces back to one 2 AM decision to click on a random YouTube video instead of sleeping.