ai-job-search

Product Introduction

1. Definition: AI Job Search is an open-source, local-first automation framework for the job application process. Technically, it is a structured workflow built on top of AI agent platforms like Claude Code, designed to function as a full-stack job application assistant that runs entirely on a user's local machine.
2. Core Value Proposition: It exists to enable job seekers to apply efficiently at scale while maintaining personalized, high-quality applications, all within a private and secure environment. Its primary value is automating the time-consuming tasks of job discovery, fit evaluation, CV tailoring, and cover letter writing using large language models (LLMs), freeing the user to focus on strategic networking and interview preparation.

Main Features

1. Intelligent Job Scraping & Ranking: The framework includes a suite of CLI tools (built with Bun/TypeScript) that scrape multiple job portals simultaneously. It features a /scrape command that searches portals like Jobindex, LinkedIn, and Jobnet, deduplicates results, and presents them. The /rank command then uses parallel AI agents to batch-score each posting against a structured fit framework (skills, experience, culture, location, career alignment), producing a ranked shortlist with identified strengths and gaps.
2. Drafter-Reviewer Application Pipeline: The core /apply command implements a multi-agent workflow. A drafter agent evaluates a job posting, tailors the user's CV and cover letter from LaTeX templates, and compiles them to PDF. A separate reviewer agent then critiques the drafts, researches the company, and suggests improvements. This iterative process ensures high-quality, context-aware application materials.
3. Local LaTeX Compilation & ATS Verification: Unlike cloud-based resume builders, this tool uses local LaTeX distributions (lualatex, xelatex) for PDF generation. It includes a verification loop that compiles, visually inspects, and iteratively fixes PDF layout issues (e.g., page breaks, fonts). It optionally uses pdftotext to extract the PDF's text layer and perform an Applicant Tracking System (ATS) compatibility check, verifying contact details and keyword coverage against the actual text a parser would read.
4. Comprehensive Application Lifecycle Management: The framework tracks the entire job search funnel. The /outcome command archives submitted materials and records interview stages, offers, and rejections. The /interview command generates stage-specific preparation packs using archived data and company research. The /html-report command creates an offline dashboard with funnel analytics from the job_search_tracker.csv.
5. Profile-Centric Skill Gap Analysis (/upskill): This feature analyzes the discrepancy between the user's profile and either a target job or an aggregate of seen postings. It produces a prioritized heatmap of skill gaps and generates a concrete learning plan with web-searched resources and time estimates, turning job search data into actionable career development insights.

Problems Solved

1. Pain Point: The immense time cost and mental fatigue of manually tailoring a unique CV and cover letter for every job application, especially when applying at scale.
2. Pain Point: The risk of submitting application documents with poor formatting, ATS-unfriendly elements, or layout errors that are only visible in the final PDF, damaging a candidate's professional image.
3. Pain Point: The difficulty in systematically evaluating a high volume of job postings to identify the best fits based on a personalized set of criteria beyond just keywords.
4. Target Audience: Technical professionals (software engineers, data scientists, AI engineers, DevOps) and knowledge workers who value automation, privacy, and data ownership. It is particularly suited for active job seekers in competitive markets.
5. Use Cases: A professional conducting a confidential job search who cannot use cloud-based SaaS tools. A candidate applying to 50+ positions who needs consistent, high-quality materials for each. A user wanting to derive a skill development roadmap from their unsuccessful applications.

Unique Advantages

1. Differentiation: Unlike SaaS platforms like Teal or Jobscan, AI Job Search is self-hosted and private; all data (profile, documents, applications) remains on the user's machine. Unlike generic AI chatbots, it provides a structured, repeatable workflow with enforced rules (e.g., honesty rule, drafter-reviewer, PDF verification) rather than a single, unpredictable conversation.
2. Key Innovation: The integration of a mandatory PDF compilation and inspection loop within the AI workflow. This technical step ensures the final deliverable is not just correct in code (.tex file) but also in its rendered form (PDF), automatically fixing common LaTeX rendering issues that break resumes. This bridges the gap between AI-generated content and professional, submission-ready output.

Frequently Asked Questions (FAQ)

1. Is AI Job Search safe to use with my private data? Yes, the core framework runs locally on your machine. Your profile, CV, and application history are never sent to a remote server unless you explicitly use optional sync features like /gmail-sync or /notion-sync, which use official, permission-scoped APIs.
2. Can I use AI Job Search with AI agents other than Claude Code? Yes. The portal search CLI tools (/.agents/skills/) work with any agent tool that can execute shell commands. The full application workflow is documented in AGENTS.md for adaptation to platforms like Cursor, Windsurf, or GitHub Copilot.
3. Do I need to know LaTeX to use this tool? No. The framework provides working LaTeX templates (cv/, cover_letters/). The /apply command handles all LaTeX tailoring and compilation automatically. However, for deep customization, basic LaTeX knowledge is beneficial.
4. How does the ATS check work, and is it accurate? The ATS check uses pdftotext to extract the raw text from your compiled CV PDF, simulating what a parser sees. It checks for readable contact info and scores keyword coverage from the job description. Accuracy depends on the PDF's text layer quality; the tool's iterative LaTeX fixes aim to maximize this. It is a robust sanity check, not a guarantee of perfect ATS parsing.
5. What if the job portals for my country aren't supported? The framework includes an /add-portal command that guides you through creating a new job-portal search skill. It investigates the portal's search patterns and scaffolds a new CLI tool following the same structure as the provided Danish portals, making it adaptable to any job market.