Method
Embedding similarity matching
Cosine-similarity matching of embeddings for dedup, retrieval or classification.
65 cases from 42 companies apply Embedding similarity matching. 17 were checked in an online A/B test; 13 report results in production.
Typical tasks: Learn user and item embeddings, Retrieve and rank from millions of items, Recommend to new users and new items, Rank search results and 15 more. Companies: DoorDash, Airbnb, Netflix, Ramp, Yelp, Pinterest and 36 more. Often combined with: LLM-as-a-judge, Gradient boosting, Multimodal embeddings, Human in the loop, RAG and 121 more.






