Job-Board Automation Target Finder

What it is

A pipeline (scrape, enrich, classify, report) that collects postings from a remote-jobs board, scores each against 14 automation categories with weighted regexes, and sends flagged postings to Claude for a structured verdict: duties it would automate, duties kept human, what Black Lily would build and a one-line outreach hook. The current data folder holds 3,000 scraped postings and 437 cached Claude verdicts, and the report stage produced roughly 230 target companies grouped by category, with a Smartlead-ready company export.

The problem

Finding companies that are hiring for work an automation could do, and knowing what to pitch each one, is slow manual research.

How it works

1. Playwright records the board's own JSON responses where listings render client-side, and detail pages are read from JSON-LD JobPosting blocks with polite threaded fetching.
2. taxonomy.py blocks roles software cannot replace and flags manager titles so they are pitched capacity instead of replacement.
3. Claude returns structured JSON per posting; verdicts are cached in JSONL and the model is configurable (a stronger model for quality, Haiku for volume).
4. The report stage groups by normalized company, sorts staffing agencies last and writes Markdown and CSV outputs.

Stack

Disciplines

For your business

AI-driven lead generation that turns public hiring signals into a ranked list of companies with a tailored automation pitch for each.