runtimelab▮
Sign in

~/runtimelab $ cat ./semantic-search/README

Build your own Semantic Search

The retrieval system around the index, not the index itself — chunking, query and document asymmetry, BM25, reciprocal-rank fusion and reranking, scored with recall@k, nDCG and MRR. It calls a nearest-neighbour index; vector-db is the course that builds one.

what you end up with
A hybrid retrieval pipeline you can score, not guess at

stages
0
level
Advanced
track
AI/ML Engineering
language
Python
  • chunking, and the answer cut in half by a boundary
  • query and document asymmetry
  • BM25 — IDF, saturation, length normalisation
  • reciprocal-rank fusion over disagreeing rankers
  • reranking, and paying for it in latency
  • recall@k, nDCG and MRR on a labelled set
  • a stale index, and knowing that it is stale
  • failure analysis instead of a nearest-neighbour eyeball
Sign in to startBrowse all challenges

The stage list opens once you sign in — it is the challenge, not the pitch. Signing in is free, and the first stages of every challenge are too.