Solving Cold-Start with Multimodal Embeddings: Netflix's MAPS System

Explain how multimodal embeddings (CLIP, MediaFM) solve the recommendation cold-start problem, and why content-aware embeddings and model consolidation had to be combined — not tested separately — to produce a real production win.

4 sections · 9 lessons

Course outline

The Cold-Start Problem in Recommendation Systems

  1. Why ID-based models fail on new items
  2. What CLIP embeddings are and why they transfer taste signals

From Content-Awareness to Model Consolidation

  1. How CLIP embeddings enable merging per-canvas models into one
  2. Reward-based weighting to balance uneven data sources

Why Two Fixes Failed Alone but Won Together

  1. Ablation testing and the second blocking factor trap
  2. Real-world stress test: absorbing a sudden distribution shift

Reusing Embeddings Across Search, Video, and Infrastructure

  1. Query-aware ranking reusing the same embeddings for free
  2. MediaFM and multimodal video embeddings
  3. The Embedding Store: decoupling foundation models from downstream models

Start learning with Wondering