AI Job Search

An open-source project by Mads Lorentzen (MadsLorentzen/ai-job-search).

What it does

A job search is a loop of repetitive work — re-tailoring a CV for each posting, writing a fresh cover letter, tracking applications, and prepping for interviews. Doing each step by hand is draining, and optimizing a CV against a posting means starting over every time.

AI Job Search bundles that flow into Claude Code slash commands. It requires Claude Code to run, and it orchestrates a multi-step workflow — from evaluating postings to tailoring CVs, writing cover letters, and interview prep — inside the session. Your data and documents stay on your machine. From the README: “The job search that runs on your machine.”

What you can make

Running the commands produces these outputs, per the README:

• Tailored CV PDF — /apply <url> reworks your CV against the posting. moderncv (banking style), lualatex-compiled, exactly 2 pages, with relevance-weighted content cutting when it overflows.
• Cover letter PDF — a custom cover.cls class, xelatex-compiled, one page with a visible signature. A drafter-reviewer flow drafts it, then a separate-context reviewer agent researches the company and critiques the draft before revision.
• Fit-ranked job shortlist — /scrape and /rank search portals, deduplicate, and score by fit against your profile, returning a ranked shortlist with gap analysis.
• Interview prep package — /interview maps likely questions to your documented STAR examples, researches company intel, and runs a mock-interview roleplay.
• Skill-gap learning plan — /upskill lays out skill gaps versus tracked postings as a heatmap and produces a learning plan with resources and time estimates.

The flow: /setup to build your profile → /scrape to collect and rank jobs → /apply <url> to generate the CV and cover letter → /interview to prep → /outcome to record results.

Key features

• Slash-command workflow

  Run each stage as a command — /setup, /scrape, /apply, /rank, /interview, /outcome, /expand, /upskill. /expand enriches your profile by scanning GitHub repos, portfolio, and Kaggle.

• PDF and ATS verification

  /apply compiles the CV and cover letter, then inspects the PDFs to catch orphaned job titles and spilled pages, and extracts the text layer to confirm contact-detail legibility and ATS keyword coverage.

• Portal scraping

  Supports Danish portals (Akademikernes Jobbank, Jobindex, Jobnet) plus LinkedIn (public endpoints) and freehire. Use /add-portal to generate a skill for your own local job board.

• LaTeX templates

  Keeps CVs and cover letters as LaTeX, and /add-template registers custom templates while preserving structure and ATS readability.

Usage

Install — fork and clone the repo, then install the portal CLI tools with Bun.

gh repo fork MadsLorentzen/ai-job-search --clone
cd ai-job-search
for tool in jobbank-search jobdanmark-search jobindex-search jobnet-search linkedin-search freehire-search; do
  cd .agents/skills/$tool/cli && bun install && cd ../../../..
done

Setup and run — start Claude Code, build your profile with /setup, then use the commands.

claude
/setup
# then /scrape, /apply <url>, /interview, etc.

Notes

• Requirements — Claude Code CLI, Python 3.10+, Bun, and a LaTeX distribution with lualatex and xelatex (TeX Live, MacTeX, TinyTeX, or MiKTeX). ATS text-layer verification optionally uses poppler’s pdftotext and degrades gracefully if it’s absent.
• Portal coverage — the scrapers center on Danish job boards plus LinkedIn and freehire. The CV, cover letter, and interview features themselves are region-agnostic.
• Fork and own — it’s not distributed via a plugin marketplace; you fork the repo and edit or extend it yourself.
• MIT License — published as open source.