George Buck

Cold-start music discovery app that learns taste from quick like/dislike reactions to song previews. Built to evaluate whether music foundation-model embeddings (MERT) outperform metadata baselines under sparse feedback.

I'm a Data Science student at NYU interested in applied ML, unconventional solutions, and making the world a brigher place. I like building things that lie in unusual intersections, are technically interesting, and sometimes even useful!

Right now I'm focused on music recommendation — specifically whether audio foundation-model embeddings can beat metadata-only baselines when a user has given you almost nothing to work with.

Previously I've worked on [add your background here].