AI Playlist Curator

Problem

Algorithmic playlists on streaming platforms are decent but generic. They optimize for engagement metrics, not for the specific context you’re in — whether that’s deep focus work, a particular creative mood, or discovering music that connects to what you already love in non-obvious ways. I wanted to see if I could build something more intentional.

Approach

Built a Python-based tool that pulls listening history and track audio features from the Spotify API, then uses clustering and lightweight ML models to identify patterns in what I listen to and when. The system groups tracks by mood and energy profiles rather than just genre, then generates playlists tailored to specific contexts.

Current capabilities:

• Listening pattern analysis across time of day and activity
• Mood-based clustering using audio features (energy, valence, tempo)
• Context-aware playlist generation
• Discovery mode that finds tracks similar in feel but outside usual genres

Outcome

This is still very much an experiment. The mood clustering works surprisingly well — better than I expected for a relatively simple approach. Discovery mode surfaces interesting finds regularly. Next steps are adding natural language input (“playlist for late night coding”) and exploring whether this could be useful to others beyond my own listening habits.