Case study: how LinkedIn rebuilt search with semantic embeddings and 2-stage ranking
_jaydeepkarale · x · 2026-11-06
A full teardown of LinkedIn's AI-powered content search: keyword matching failed by returning nothing or irrelevant posts, so LinkedIn added semantic capability. A token retriever and a two-tower embedding retriever narrow billions of posts to thousands (post embeddings precomputed offline), then L1/L2 rankers—MLP models scoring on-topicness and long-dwell—combine multi-objective scores to produce the final ranked feed. Likes and comments are excluded as quality signals since they can be farmed.
More from Infra
- LMSYS XM-Sparse cuts decode time up to 63.6% with HiSparse plus X-Mem KV offload — xiaosun86 · 2026-11-06
- Polymarket poll: 56% of voters say AI and data centers shaped their midterm vote — KyeGomezB · 2026-11-06
- New attention kernels meet real models: where interpretability overlaps systems research — QuentinAnthon15 · 2026-11-06
- MoE All-to-All Comms Eat 60% of Training Steps; Correlated Placement Cuts 58% of Dispatched Rows — QuentinAnthon15 · 2026-11-06
- Texas data center bill would let all permitted projects finish, just with stricter consumer rules — VraserX · 2026-11-06
- GitHub rebuilds Git storage for thousands of concurrent AI agents — arpit_bhayani · 2026-11-06