Outline sketch
Audience: recruiting ops / ATS.
Steps:
1) Parse resume into structured state + role must-haves.
2) Per requirement: Score/Choice; Noul red_flags.
3) Code ranks candidates; low confidence → human review.
Expected effect: consistent first pass.
Ban: protected-class or demographic scoring.
Needs access to: typesafe-sdkats
Who it's for
Recruiting ops / ATS integrators
Steps / how it's set up
Keep high-stakes decisions with humans. Sources: TypeSafe use-case map · recruiting cluster notes in awesome-jev-usecases · community CV entries in cobanov/awesome-jev (label community).
Prefer must-have Scores for hard gates (degree, years, location policy). Keep protected-class attributes out of prompts and scoring inputs; document legal review. Thresholds are not universal — label any published numbers as author-reported.
Sources (wave-2 deepen)
Expected effect
Same closed-label + confidence pattern as triage, applied to must-have checks.
Unofficial outline for learning. Paraphrased from public docs and tutorials — not a production recipe. Review sources before you automate anything.