career-ops-3.0

An AI job-search pipeline with a coded LangChain engine. It scans job boards, fetches descriptions, filters by experience, rates every job against multiple resumes, and generates a heavily-tailored resume per job — choosing which of your GitHub projects to feature and rewriting skills/keywords to match the JD.

Lineage: career-ops-3.0 builds on the open-source MIT project career-ops by @santifer (the scanning core, tracker, and dashboard). The intelligence layer here is reworked from agent-driven prompts into a coded LangChain.js pipeline (lc/), and the resume engine descends from the CareerOS pattern. See LICENSE for attribution.

LLM backend

The LangChain layer (lc/) runs on Ollama (local, free, unlimited) by default, or Gemini (cloud). Switch with LLM_PROVIDER in .env.

• Ollama: install from then ollama pull llama3.1.
• Gemini: set LLM_PROVIDER=gemini and add GEMINI_API_KEY (free tier is

~20 requests/day per model — fine for testing; enable billing for volume).

Dashboard

The tracker schema is unchanged; ratings / best-resume / min-experience come from the data/eval-meta.json sidecar via /api/eval-meta and the enriched /api/daily.

Ethics

Tailoring reframes and reorders your real experience and projects to match a JD — it never invents skills, employers, or metrics. Quality over quantity; always review before applying.

License

MIT — see LICENSE.