Browse / Job hunt / ai-job-searcher
ai-job-searcher
ymuromcev
Multi-profile job search pipeline.
Open ymuromcev/ai-job-searcher → Checked against the four rules.
Built for Claude Code
ships .claude CLAUDE.md and nothing for other agents.
In its own words
One engine, multiple candidates.
A personal AI agent for job search. One engine, many profiles. Built on Claude Code + a plain Node.js toolkit (no build step); all the AI lives behind the CLI.
What it is
A job-search pipeline that scans ATS feeds and job boards, filters roles against my preferences, drafts tailored cover letters, pushes everything into Notion as a Kanban pipeline, and reads my Gmail to move cards when recruiters reply.
I built it for myself. When a family member also needed a job search, I rebuilt it so one codebase could serve multiple candidates without code duplication.
Evolution
It started as a single-user Python + AppleScript Frankenstein — hardcoded to me, hardcoded to one Notion workspace. After ~1 month of daily use and a clear picture of what actually mattered, I rewrote it from scratch:
- v1 (prototype) — single-user, monolithic. Shipped fast, was
good enough to land interviews, accumulated drift as fast as I shipped features.
- v2 (this repo) — engine cleanly split from profile data, so one
deployment serves multiple candidates. Per-profile Notion DBs, per-profile resume archetypes, per-profile salary models. Onboarding a new candidate is a 15-minute conversation with Claude (the onboard-profile skill drives the questionnaire), not a day of copy-paste.
The old v1 codebases are not public — they predate the clean-slate rewrite and carry personal data in their history.
The repository's own description, then 16% of its README. Read the rest on GitHub ↗
Read from the repository's git tree on 2026-08-17.