fun-rec

Why we included this project

This book-style repo walks you through recommender systems the way a course would: collaborative filtering, vector and sequence recall for candidate generation, feature crossing and multi-objective modeling for ranking, then reranking for diversity. The second half is where it stands out, covering generative paradigms like LLM-based modeling, scaling-law architectures such as HSTU, end-to-end designs inspired by OneRec, reasoning-enhanced recommenders, and diffusion-based models, then closing with a production-grade build project. It's aimed at readers who already have some machine learning background and want either to prepare for interviews or to plan a real recommendation system from scratch; the mix of theory, TensorFlow code, and a worked full-stack example serves both. Since the project is still evolving and updates often, treat it as a companion text rather than something you'd drop into production.