Job Application Agent

An AI agent that does the whole job-application loop: discovers open roles, tailors your resume per job description, generates a PDF, and fills out the ATS application form (Greenhouse, Ashby, Lever, ...) — dropdowns, comboboxes, and the "why do you want to work here" boxes included. Then it stops and waits for you to review and click Submit. It never submits on its own.

▶️ Watch the full demo — it applies to real jobs on camera

What it does

• Discovers open roles across job boards (discover.py, yc_discover.py)
• Tailors your resume to each job description and generates a fresh PDF (tailor.py, gen_pdf.py, make_cv.py)
• Fills the actual application in its own browser window — so your main browser stays free, unlike extension-based autofill (main.py)
• Pauses for human review before every submit — it's an assistant, not a spam cannon
• Survives crashes: queue processing picks up where it left off
• Follows up: drafts follow-up emails for submitted applications (followup.py)

How it works

1. profile.yaml holds your facts + voice notes (single source of truth)
2. Playwright opens the job URL in a dedicated Chromium with a persistent profile (logins/cookies survive between runs)
3. The form is extracted from the DOM; an LLM maps every field to an answer — profile facts verbatim, open-ended questions drafted in your voice, unknowns skipped
4. The agent fills the form, then pauses — you review and click Submit yourself

Setup

python -m venv .venv && source .venv/bin/activate
pip install -r requirements.txt
playwright install chromium
# create profile.yaml with your details (name, contact, work history, voice notes)
export LLM_API_KEY=...         # and optionally LLM_BASE_URL / LLM_MODEL

Run

python main.py https://boards.greenhouse.io/acme/jobs/12345
python main.py --queue jobs.txt    # one URL per line
./apply.sh                          # full pipeline via launcher script

Notes

• Quality over volume: review every application before submit
• Don't point this at LinkedIn Easy Apply (ToS) or CAPTCHA-walled flows
• Workday: log in once in the agent's browser window; the persistent profile remembers
• Browser profiles, resumes, and application data are gitignored — keep it that way; they contain session cookies and personal info