Details
The system ingests terabytes of daily behavioral signals — including viewing duration, pause patterns, skip behavior, search queries, and device context — and combines collaborative filtering, deep learning, and graph-based models to generate ranked recommendations at sub-100ms latency for each user. Netflix's internal 'Hydra' multi-task model handles homepage ranking, search result ordering, and notification personalization simultaneously, using shared user and content representations. Netflix has publicly confirmed that approximately 80% of content viewed on the platform is driven by this recommendation system.
Have evidence about Netflix's AI practices? Submit a report.
Report a Sighting →