Parser agents normalize job descriptions and CVs into one schema, hybrid retrieval builds a shortlist, and eligibility, competency and differentiator agents score each pair under config-driven rules. Every score ships with its evidence, and recruiters stay in the loop.
FOR YOU
You match people to roles at volume and need rankings that recruiters can trust, audit, and override, not a black-box score.
§ CAPABILITIES
Multi-agent scoring01
Hybrid retrieval02
Explainable ranking03
Human-in-the-loop04
Config-driven rules05
Cost and quality telemetry06
§ IN DETAIL
a.
Parser agents normalize job descriptions and CVs into one schema, cached for repeat runs
b.
Hybrid retrieval (embeddings + BM25) builds the shortlist before any LLM scoring
c.
Eligibility, competency and differentiator agents with config-driven gates, caps and floors
d.
Flags instead of filters: weak matches are explained, not silently dropped
e.
"Why this score?" evidence trace and full run history for every match
f.
A 960-pair scoring audit (733 ranking inversions) drove a rebuilt scoring model