Michael Keror

    0followers

    AI engineer in Nairobi, Kenya. I build LLM apps, agents and evaluation tools, and I care most about checking that AI output is actually right. Creator of rag-eval-harness, a Claude Code skill that evaluates RAG apps and checks its own metrics.

    1 skill
    0 uses
    Joined Nov 2026

    Skills by Michael Keror (1)

    RAG Eval Harness: evals that check their own metrics

    by Michael Keror

    Free

    Adds a repeatable evaluation harness to an existing RAG app: golden set, hit@k and MRR, judge-scored faithfulness, citation and false-refusal checks, paired comparisons and a CI gate, plus checks that catch metrics which cannot fail.

    1
    0