Drop Kai into Claude and get a senior RAG architect who builds retrieval that returns the right context, measured with an eval harness before you tune. Kai treats retrieval as an information-retrieval problem first: corpus prep and chunking strategy, embedding selection, vector index tuning (HNSW/IVF), hybrid search (BM25 + dense) with RRF fusion, cross-encoder reranking, query transformation (multi-query, HyDE, decomposition), retrieval and end-to-end evaluation (recall@k, MRR, nDCG, faithfulness), grounding and citation, freshness, and the cost and latency of every hop. Measure before you tune. Always evaluate on a held-out labeled set in dev/staging before serving in production. What you get →Chunking strategy and metadata design →Hybrid search (BM25 + dense) with RRF and cross-encoder reranking →Query transformation: multi-query, HyDE, decomposition →Retrieval eval: recall@k, MRR, nDCG, faithfulness, with before/after 📄 kai-rag-retrieval-architect.skill Under 2 min install Works with Claude, ChatGPT & any AI chat How to install Download the .skill package → open Claude → paste SKILL.md into your Project Instructions or system prompt → describe your requirement → Kai builds the answer. Includes a full worked example so you see exactly what you get.
