Project
Roadmap
Scope is fixed and the cut list is real: a smaller product that is beautiful, tested, and honest beats a large one that is half-wired.
Shipped — on main today
- Full scoring engine: rubric → per-criterion scores with quoted evidence spans at exact offsets
- Review workflow: ranked queue, required-reason decisions, blind review, score overrides
- Adverse-impact bias audit with the four-fifths rule and opt-in self-reported demographics
- Tamper-evident hash-chained audit log of every run, override, and decision
- Semantic search & ask over resumes (pgvector) with server-verified citations
- Rubric editor: criteria, weights, anchored scales, versions, AI draft
- Interview kits: printable debrief packs with probes, anchors, and the evidence behind every score
- Candidate portal: one-shot opt-in demographics self-report links
- Team seats: invite, roles (owner / recruiter / hiring manager / viewer), pending-invite management
- Candidates table: server-side pagination (25/page) and identity search across name, email, and phone
- Shareable scored reports (public read-only link, instant revoke, view counts) + hash-chain-verified audit CSV export
- Compare view: 2–3 candidates side-by-side on one scoring run
- Eval harness: direction accuracy, self-consistency, position bias, name-swap bias — in CI
- Public REST API with a live OpenAPI 3.1 contract; zero-dependency CLI
- One-command self-host via Docker Compose; local Ollama mode
- This premium site with a live no-signup demo running the real engine
- Skill-graph adjacency — every candidate page now shows matched / adjacent / missing skills computed deterministically from the parsed resume against skills extracted from the job description (no extra LLM cost)
- Talent-pool rediscovery — one org-wide search across every candidate from every job: identity search, skill facet chips, stage filter, rubric-weighted minimum score; shortlisting records a regular, audited decision. Blind review honored
- Public parser API — POST /api/public/parse: stateless resume → JSON-Resume-compatible profile with a Bearer org API token (file or text), SHA-256 content hash, per-token rate limiting; nothing stored
- ATS format check — deterministic parse simulation (pass/warn/fail, 0–100, actionable findings) on every upload: scanned pages, table/multi-column layouts, mojibake, missing sections — before a scoring run is spent on a mangled resume
Between here and launch
In progress — launch window (Sept–Oct 2026)
- v1.0.0 tagged with GitHub Release, CHANGELOG, and GHCR images (done)
- npx hirelens CLI published on npm (done)
- Private beta with real recruiters — feedback drives the polish list
- Launch posts: Product Hunt and Show HN (drafted in docs/launch-kit.md)
Shipped — v1.1: the candidate side, at equal level
- Candidate report — recruiters share an evidence-backed, read-only score report with each candidate via a once-only token link (create → view → read-counted → revocable)
- Score my resume — free public tool: candidates score their resume against any job description with the same rubric engine, evidence quotes included
- For-candidates hub — what transparent screening means for you, and control over your demographics data
After launch
| Version | Theme |
|---|---|
| v1.4.0 — shipped 6 Oct 2026 | ✅ talent-pool rediscovery · ✅ public parser API (token-gated, stateless) · ✅ ATS format check · ✅ parser heuristic hardening (compact resumes) · ✅ talent-pool “add to job” |
| v1.3 — Mar 2027 | i18n, SSO/SAML, advanced RBAC |
| v1.5 — 2027 | Candidate application tracker — candidates organize applications and see which of their skills matched each role |
| v2.0 | Plugin architecture, Helm chart for self-hosters, fine-tuned open scoring model |
Follow along issue-by-issue on GitHub.