JobSearch Agent
Applying to jobs by hand is slow and repetitive — exactly the kind of work I automate for clients. So I built an agent to do it for me, end to end.
The problem
A good job search means reading hundreds of postings, judging fit, rewriting your resume for each, writing cover letters, applying, and tracking responses. Done manually it caps you at a handful of quality applications a day and burns hours.
My approach
A daily pipeline fetches roles from scrape-friendly boards (Greenhouse, Lever, Ashby, RemoteOK, YC, Wellfound), filters by my rules (remote/visa, startup, AI roles), and uses Claude to score each role and explain the match. For anything above threshold it tailors a resume variant and cover letter, then queues a prepared application card. A local Playwright runner auto-fills forms so I just review and click submit, and a Gmail watcher logs every reply and drafts follow-ups.
The result
Turns a full day of manual searching into a 10-minute morning review of pre-matched, pre-written applications — with full documentation of every step and an interview-prep generator mapped to my matching projects.
Human-in-the-loop by design
Nothing is submitted silently. Pakistan-based companies are always shown first and never auto-applied. The agent prepares; I decide.
Match explainability
Every score comes with the why — which skills matched, which were missing, and what single project I could add to push a 65% role to a strong yes.
Score a job against my profile
Runs locally in your browser — a preview of the LLM match-scoring in the real agent.