Collaborative Filtering Movie Recommender & Latent Space Analysis

A collaborative filtering movie recommender built with Python using the MovieLens 32M dataset. It implements low-rank matrix factorization optimized via stochastic gradient descent with $L_2$ regularization, leverages @njit C-compilation for a 100x training speedup across 25.6M ratings, and delivers an 1100% precision increase over my previous content-based model's baselines while uncovering hidden taste clusters using t-SNE visualization.

View Project →

Content-Based Movie Recommendation Engine

A movie recommender system built with Python and Scipy using the MovieLens 32M dataset. It explores the transition from item-item to user-item recommendations, evaluates implicit vs. explicit feedback models, and demonstrates how popularity filtering boosts recommendation precision from under 1% to over 16%.

View Project →