nitish1986
Follow
I work on production readiness for AI systems: retrieval quality, evaluation, cost, and the failure modes that show up after a RAG or LLM feature ships.
Before that, 11 years building and running production software for startups. Most of the repos below are from that work.
Recall@k, nDCG and LLM-judge agreement for retrieval on BEIR subsets or your own corpus. Small, auditable, runs offline.
Python
Ways RAG systems fail in production: public repro, metric, measured fix and production caveat for each
Python
