← all projects

2025–26 · ml · backend

VibeCurator

A Spotify music recommender that generates DJ-quality track transitions across a 1.2M-track catalog.

context

Personal project

stack

Python, FastAPI, PostgreSQL, asyncpg, Spotify OAuth

what i built

  • Engineered a production-grade Spotify music recommender in FastAPI with asynchronous PostgreSQL (asyncpg) over a 1.2M-track catalog aggregated from Kaggle, ReccoBeats, and Deezer.
  • Designed a three-layer catalog architecture combining TrueSkill pairwise ranking with Camelot Wheel harmonic-mixing theory to generate DJ-quality track transitions.
  • Built RESTful API endpoints and Spotify OAuth integration to serve personalized, low-latency recommendations at scale.