Reviewing job postings was taking hours — most of them wasted

Five sources a day, most of it spent on roles that didn’t fit. A triage layer felt more useful than more tabs, so a fully deployed local Python agent now crawls three job boards directly and parses Gmail alerts for the two that block scrapers outright.

Every listing gets scored against a detailed personal rubric before a Trello card ever gets created, with a keyword pre-filter running first to keep API costs down. Rejections and interview invites move their own cards based on incoming email signals — no manual updates, ever.