Watches company job boards, scores every opening against your resume, and emails you when something worth applying to shows up.

It tracks 56 companies across seven ATS platforms — Greenhouse, Lever, Ashby, Workable, Workday, Oracle Recruiting Cloud — which is what it takes to cover both startups and large enterprises on one list.

• Finds new openings. Every job gets a stable id, so "new" means new.
• Notices when jobs close. A posting that disappears is marked dead and hidden.
• Ranks against your resume. 0–10 with a reason, from an LLM that reads the job description.
• Daily alerts for a shortlist of companies — email plus a desktop notification.
• Weekly digest every Monday, ranked.
• A local web UI to search and filter everything.

Two rules run in code, not in the prompt, because a model asked nicely will comply inconsistently:

Seniority. A Senior/Staff/Principal title caps at 3/10 however well the skills match. Without this, senior roles held 31 of the top 100 slots and the first junior role sat at position #67 — the jobs a new graduate could actually get were buried under the ones they could not.

Relevance. A role weak on both skills and domain caps at 3/10. Otherwise a branch-office internship scores 8/10 for a developer purely because it is an internship.

Both are arithmetic on data already stored, so recap re-applies them for free.

Needs Python 3.11+ and a Mac (the schedules use launchd).

git clone https://github.com/suvamneog/jobradar.git
cd jobradar
./install.sh

Then three things:

1. Your resume — put a PDF in the folder, point resume_path at it in config.yaml.

2. Your keys — cp run.example.sh run.sh, then fill in:

• An LLM key. Any OpenAI-compatible endpoint works; the default is xkiro with Mistral Large 3, which has a free tier.
• A Gmail App Password for sending mail. Needs 2FA on the account first.

3. Check it works

./run.sh test          # sends a test email
./run.sh scan          # fetch, diff, score
./run-web.sh           # http://localhost:8765

companies:
  - { name: 'Adobe', board: workday, slug: 'adobe/wd5/external_experienced' }

alerts:
  companies: [Adobe, Oracle, Atlassian]   # checked daily
  min_score: 6

llm:
  provider: xkiro
  model: mistralai/mistral-large-2512
  daily_token_limit: 1000000

Find a company's board and slug in its careers URL:

./run.sh scan          # fetch, diff, score
./run.sh alert         # check the shortlist, alert if anything opened
./run.sh digest        # send the weekly email
./run.sh recap         # re-apply the scoring rules, no tokens
./run.sh list          # the week's finds in the terminal
./run-web.sh           # web UI

install.sh registers two launchd jobs: daily alerts at 09:30, weekly digest Mondays at 09:00.

Free tiers have a daily cap, so the budget is checked before every request and recorded per batch. A run that hits the ceiling stops and resumes next time — it can't lock you out of your key.

Jobs are scored 25 per request with compact positional replies. A full 3,400-job scan costs about 340K tokens. After that only new postings are scored, which is a few thousand a week.

for t in tests/test_*.py; do ./.venv/bin/python "$t"; done

Ten suites. Most exist because something broke in use — a threading crash in the live scan, an alert that would have fired 142 notifications at once, a malformed reply that killed a 3,000-job run.

See OVERVIEW.md for architecture and design notes.

Built by @suvamneog. The grid UI took small inspiration from this post.

Scores are a sort order, not a verdict — a 4/10 is still worth a glance.

MIT