How to Configure the AI for Personalized Job Searching

You configure the AI for personalized job searching by completing a three-step onboarding process in the ai-job-search repository: populate your candidate profile in CLAUDE.md, define fit-evaluation rules in 04-job-evaluation.md, and select LaTeX templates and job portal tools before running the /setup, /scrape, and /apply commands.

The MadsLorentzen/ai-job-search repository provides a Claude Code-driven job application assistant that automates the search and application process. To configure the AI for personalized job searching, you must supply your career profile, scoring preferences, and document templates across three specific configuration files. This guide walks through the exact file paths and commands needed to personalize the system.

Step 1: Build Your Candidate Profile

The personalization system reads your career data from two locations. The master profile lives in CLAUDE.md at the repository root, while the structured template is stored in .claude/skills/job-application-assistant/01-candidate-profile.md.

Run the /setup command in the Claude CLI to launch an interactive interview. The wizard extracts data from three sources:

1. A populated documents/ folder containing PDFs of your existing CV and diplomas
2. A single pasted CV text
3. A live interview session

The wizard writes the extracted values into the skill markdown files, creating a persistent candidate profile that drives all future applications.

Step 2: Define Fit-Evaluation Rules

To control how the AI ranks job postings, modify .claude/skills/job-application-assistant/04-job-evaluation.md. This file contains a JSON-style rubric that the /apply command consumes to generate a numeric fit score and textual justification.

Edit the following sections to reflect your priorities:

• Skill match: Weight specific technical skills against job requirements
• Career alignment: Define target roles, industries, and seniority levels
• Location constraints: Set geographic preferences or remote work requirements
• Deal-breakers: List non-negotiable exclusions (e.g., specific technologies or company types)

Changes to this file instantly influence future searches, as the /scrape and /apply commands reference this rubric when evaluating postings.

Step 3: Select Templates and Job Portal Tools

The system uses LaTeX for document generation. The default CV template is located at cv/main_example.tex and uses the moderncv banking style compiled with lualatex. The default cover letter template resides at cover_letters/cover.cls and uses Raleway fonts compiled with xelatex.

To register a custom template, run /add-template and point the wizard to your .tex file. The command records compile instructions and stores the template under templates/.

For job searching, the repository ships with CLI tools for four Danish portals: Jobbank, Jobdanmark, Jobindex, and Jobnet. Install these tools with bun install in their respective directories under .agents/skills/. To add a new portal, run /add-portal and follow the guided scaffolding.

Running the Personalized Workflow

Once configured, execute the core commands to run your personalized search. The /scrape command searches all configured portals and returns postings ranked by your fit-evaluation rubric. The /apply <url> command takes a specific posting, re-evaluates fit, auto-generates tailored CV and cover letter PDFs, runs ATS-text-layer checks, and returns verified submission materials.

The entire pipeline is profile-driven—any update to CLAUDE.md or the evaluation markdown immediately affects future outputs.

# Clone the repository and install portal tools

gh repo clone MadsLorentzen/ai-job-search
cd ai-job-search

cd .agents/skills/jobbank-search/cli && bun install && cd ../../..
cd .agents/skills/jobdanmark-search/cli && bun install && cd ../../..
cd .agents/skills/jobindex-search/cli && bun install && cd ../../..
cd .agents/skills/jobnet-search/cli && bun install && cd ../../..

# Launch Claude Code and run the interactive setup

claude
/setup

# (Optional) Fine-tune the evaluation rubric

nano .claude/skills/job-application-assistant/04-job-evaluation.md

# Execute the job search and application workflow

/scrape
/apply https://jobindex.dk/job/1234567

# Add custom templates or portals as needed

/add-template
/add-portal

Summary

• Configure your candidate profile in CLAUDE.md and .claude/skills/job-application-assistant/01-candidate-profile.md using the /setup wizard
• Personalize fit-evaluation rules in .claude/skills/job-application-assistant/04-job-evaluation.md to control how the AI scores job postings
• Select LaTeX templates (cv/main_example.tex, cover_letters/cover.cls) and job portal tools (Jobbank, Jobdanmark, Jobindex, Jobnet) for your target market
• Execute /scrape to search and /apply to generate tailored applications based on your configuration

Frequently Asked Questions

What file contains the candidate profile configuration?

The master candidate profile is stored in CLAUDE.md at the repository root, with a structured template maintained in .claude/skills/job-application-assistant/01-candidate-profile.md. The /setup command populates these files based on your documents or interview responses.

How do I customize how the AI scores job fit?

Edit .claude/skills/job-application-assistant/04-job-evaluation.md to adjust the JSON-style rubric for skill matching, career alignment, location constraints, and deal-breakers. The /apply command uses this file to generate fit scores and justifications for each posting.

Can I use my own LaTeX templates for CVs and cover letters?

Yes. Run /add-template to register custom .tex files. The system stores these in templates/ and records whether to use lualatex or xelatex for compilation. The default templates are cv/main_example.tex (moderncv style) and cover_letters/cover.cls (Raleway fonts).

Which job portals are supported by default?

The repository includes CLI tools for four Danish portals: Jobbank, Jobdanmark, Jobindex, and Jobnet. Install these with bun install in their respective .agents/skills/ directories. You can add additional portals using the /add-portal command.

{
  "mcpServers": {
    "instagit": {
      "command": "npx",
      "args": ["-y", "instagit@latest"]
    }
  }
}