GenRec: Towards LLM-Native Recommendation at Netflix

Netflix TechBlog describes GenRec, a new recommendation engine powered by large language models. Rather than relying on thousands of hand-crafted features, GenRec verbalizes user histories and content metadata as natural language, then uses a post-trained foundation LLM to rank items in the catalog. The system employs a two-phase approach: first adapting an open-source LLM on Netflix-specific data, then training it further on ranking-specific objectives and reward signals. In A-slash-B testing against Netflix's mature production system, GenRec matched or exceeded performance while using significantly fewer labeled examples and input signals—a shift from feature engineering to what Netflix calls context engineering.

Listen to this story

Hear this and more stories in a personalized audio briefing.

Open The Chonkerton