A job-search radar that watches company career pages for you. Official site: searchsteward.com
SearchSteward is a proactive job-search tool. Instead of scrolling job boards, it monitors 40,000+ company career pages directly and scores every new opening against your résumé the day it's posted — so well-matched roles find you, not the other way around.
This repo is the developer face of the product: what it looks like, how it's built at a high level, and the parts of the engine we open-source.
- Career-page radar — tracks tens of thousands of employer career pages daily and surfaces new, on-target roles as they appear.
- AI résumé ↔ role scoring — every opening scored against your background, with the matched keywords and the why shown, not just a number.
- Ghost Job Detector — flags stale, perpetually-reposted, or never-filled listings at the source, so you don't waste applications. Because we pull straight from company career pages — not aggregators, where ghost jobs pile up — we know when a listing first appeared and how it left.
- Application tracking & insights — keep every role and its status in one place, and see where your funnel actually stalls.
Matches, scored and explained. Every role gets a fit score against your résumé and search settings, with lanes for your target companies:
Every score shows its work. Click into any match: the score breakdown, matched keywords, ranking rationale — and the ghost-posting check — are all visible:
Your funnel, diagnosed. Insights shows conversion at each stage, which résumé version is out-performing, and where applications are going stale:
The front door:
We open-source the pieces where transparency is the feature. Both are MIT-licensed, dependency-free, and are the same code running in production.
resume-match/ — Résumé ↔ JD matcher + ghost-posting checker (TypeScript)
The engines behind our free, no-signup tools. Both run entirely in your browser — pasted text never leaves the page, and publishing the source is how we back that claim up.
scoreMatch()— keyword-coverage scoring with stemming, acronym handling, phrase boosting, and JD-filler suppression. Powers searchsteward.com/tools/resume-match.checkGhostJob()— a weighted checklist of documented scam/ghost-posting signals with human-readable explanations. Powers searchsteward.com/tools/ghost-job-checker.
ghost-signal/ — Production ghost-job signal (Python)
The actual decision logic behind the in-app ghost-job badge: pure, I/O-free,
and deliberately conservative. Absence of evidence yields unknown, never
clean; every tier carries its reasons; and market-universal noise (like a
missing salary) is recorded but never counted against a listing. The README
documents every signal and weight.
A single product, three layers:
- Ingestion (Python). A scraping engine with per-ATS adapters (Greenhouse, Lever, Workday, Ashby, SmartRecruiters, Workable, and many more) sweeps company career pages on a continuous cycle, normalizes listings, and tracks each posting's lifecycle — first seen, changed, closed. Board health is monitored and self-healing: moved or renamed career pages are rediscovered rather than dropped.
- Matching (Python). A two-stage scoring funnel: fast hard gates (location, disqualifiers) followed by a weighted relevance model tuned per user from their résumé and search settings. LLM analysis runs on top for fit narratives, résumé tailoring, and negotiation prep.
- Product (TypeScript). A Next.js/React app backed by a FastAPI service and PostgreSQL.
The scoring weights, gate tuning, ATS adapter details, and company registry are the proprietary core and stay closed — what we open-source instead are the user-facing checks where seeing the logic is the point.
- Résumé ↔ Job-Description Matcher — https://searchsteward.com/tools/resume-match
- Ghost Job Detector — https://searchsteward.com/tools/ghost-job-checker
- Website: https://searchsteward.com
- LinkedIn: https://www.linkedin.com/company
- X: https://x.com/SearchSteward
- Crunchbase: https://www.crunchbase.com/organization/searchsteward
Built for active job seekers who want signal over noise.