AI Data Architecture

AI Data Architecture

How Netflix Scaled Multimodal Embeddings to Personalize Content for 200 Million Users

Netflix detailed the architecture behind MAPS, their production-ready multimodal asset personalization system. By deploying foundation models like CLIP and their in-house MediaFM, MAPS solves the cold-start problem by generating personalized visual previews at massive scale....

AI Data Architecture

How Netflix solved the recommendation cold-start problem using multimodal embeddings

Netflix engineers deployed MAPS (Multimodal Asset Personalization at Scale) to eliminate the recommendation cold-start problem. By generating embeddings directly from video and audio, the system personalizes homepage assets for users the moment a new title launches....