Revibe

an ml-powered music recommendation engine over a 170k+ song dataset. uses k-means clustering as a pre-filtering mechanism to reduce the search space from 170k songs to cluster-local subsets, then ranks by euclidean distance across 9 audio features (acousticness, danceability, energy, etc). integrates itunes api for 30-second audio previews and metadata enrichment. features client-side session caching, 300ms debounced search, and a singleton audio playback controller.