Braun Board

Tell us the sales job you would take. We will tell you when it shows up.

A watchlist-driven job board built only for B2B sellers. Fresh inventory pulled from public employer ATS boards (Greenhouse, Lever, Ashby), normalized into a seller-native taxonomy, with email alerts when a job matching your watch appears.

Full product spec: docs/SPEC.md

Stack

• Next.js 15 (App Router, TypeScript, Tailwind v4)
• Supabase (Postgres + Auth) for data, accounts and row-level security
• Anthropic API for seller-specific attribute extraction (optional, falls back to keyword rules)
• Resend for alert email (optional, dry-runs without it)
• GitHub Actions cron for the 4-hour refresh and daily digest

Getting started

npm install
npm run dev        # http://localhost:3000, shows sample jobs until Supabase is connected
npm test           # classifier, comp parsing, location, match scoring, freshness

Connect real data

1. Create a Supabase project. Run the files in supabase/migrations/ in order in the SQL editor (or supabase db push).
2. Authentication setup:
   • Providers: Email is on by default. For Google, create an OAuth client in Google Cloud Console and paste the client ID/secret into Supabase under Authentication > Providers > Google.
   • URL Configuration: set Site URL to your domain and add http://localhost:3000/auth/callback and https://<your-domain>/auth/callback to Redirect URLs.
3. cp .env.example .env.local and fill in the keys.
4. Seed companies and detect their ATS:

   npm run seed -- data/seed-companies.csv
   

5. Pull jobs:

   npm run ingest -- --all
   

6. For scheduled runs, add the same env vars as GitHub repo secrets. .github/workflows/ingest.yml runs ingestion every 4 hours and the digest daily.

How ingestion works

Company → detect ATS → adapter → raw jobs → sales classifier → normalize
       → extract sales attributes → dedupe → store → match against watches → alerts

Piece	File	Notes
Adapters	src/lib/ingest/adapters/*	One per ATS, all return the same RawJob shape
ATS detection	src/lib/ingest/detect-ats.ts	Sniffs careers URLs, or probes public APIs with guessed board tokens
Sales classifier	src/lib/ingest/classify.ts	Rule-based. Includes/excludes per spec section 3, maps to role family, seniority, segment
Location	src/lib/ingest/location.ts	US-only filter, remote/hybrid/onsite. Bare "Remote" is kept but flagged needs_review
Compensation	src/lib/ingest/compensation.ts	Trust rule: only explicitly published, explicitly labeled pay. Unlabeled ranges are dropped
AI extraction	src/lib/ingest/extract.ts	Motion, self-sourcing, customer segment, summary, and what the posting does/doesn't disclose. Enum-validated, Unknown over guesses
Dedupe	src/lib/ingest/dedupe.ts	company + normalized title + location, then description similarity
Pipeline	src/lib/ingest/pipeline.ts	First-seen/last-seen, two missed crawls before a job is hidden, failed crawls never remove jobs
Match score	src/lib/match.ts	Transparent, based only on stated preferences. Unknowns are neutral. Dealbreakers block
Alerts	src/lib/alerts/*	Matches recorded at ingest, sent instant or daily, click tracked via /r/[id]

Build status (spec section 20)

• 1. Supabase schema (auth tables + RLS included)
• 2. Company seed import + ATS detection
• 3. Greenhouse adapter
• 4. Sales-role classifier + normalized taxonomy
• 5. Job grid, job cards, company page, job detail page
• 6. Lever + Ashby adapters
• 7. Dedupe, first-seen/last-seen, inactive-job handling
• 8. Watch creation + preference storage: onboarding, "Watch this search", company watches, saved jobs, sign-in (email link + Google)
• 9. Match scoring, shown on cards and job pages with the per-criterion breakdown
• 10. Email alert generation + delivery
• 11. Admin dashboard (data is there: crawl_runs, needs_review, is_disabled)
• 12. Scale from 200–500 test jobs toward ~5,000. Grow data/seed-companies.csv toward ~1,500 companies

How accounts work

• Browsing never needs an account.
• Onboarding (/for-you) works signed out. Answers are kept in a cookie so For You shows matches right away.
• Signing in (email link or Google) adopts those answers as the user's primary watch, then unlocks saving jobs, watching companies and searches, and email alerts.
• Alerts (/alerts): Instant, Daily Digest or Off, per watch. Watched companies go in the daily digest.

Notes

• Never scrape LinkedIn or anything behind auth. Every adapter hits a documented public endpoint.
• data/seed-companies.csv is a starter list of 40. Rows without a careers_url get their ATS detected by probing.
• Sample data in src/lib/sample-data.ts uses made-up companies and only shows when Supabase isn't configured.