career-ops-automation

An autonomous, deterministic job application automation system designed for AI & Machine Learning engineers. Built to replace brittle conversational agent loops with an n8n-style stateful pipeline: fail-fast invariants, persistent browser sessions, strict candidate provenance, and human-in-the-loop Mediator Mode.

⚡ Key Highlights & Core Invariants

1. No 30-Minute Agent Hangs

Agents previously drifted into endless conversational loops or hung indefinitely on complex web forms. This engine enforces hard execution timeouts:

• Maximum 90-second overall browser deadline.
• 45-second Mediator supervisor countdown.
• If user is away, it captures a diagnostic screenshot, updates SQLite to needs_manual_review, and advances to the next job in the batch without crashing.

2. Strict 2-Page Resume Policy

• All generated resumes are strictly locked to 2 pages (based on the candidate's master Awesome-CV LaTeX source).
• PyMuPDF (fitz) performs a hard runtime check: if page count != 2, compilation aborts immediately. Margins and section hierarchies are strictly preserved.