How to Configure Salary Benchmarking with Custom Data in AI-Job-Search

Place a salary_data.json file at the repository root and run salary_lookup.py to enable personalized salary comparisons during job applications.

The AI-Job-Search repository by MadsLorentzen includes a dedicated salary benchmarking system that lets you compare job offers against proprietary data. This guide explains how to configure salary benchmarking with custom data using the built-in salary_lookup.py utility and optional conversion tools.

Required File Structure

The system expects a single JSON file at a specific location:

• File: salary_data.json
• Location: Repository root (same directory as salary_lookup.py)
• Discovery path: Path(__file__).parent / "salary_data.json" in [salary_lookup.py line 27](https://github.com/MadsLorentzen/ai-job-search/blob/master/salary_lookup.py#L27)

If the file is missing, the tool exits gracefully and the /apply workflow skips salary comparison entirely.

Creating Your Custom Salary Data

Option 1: Manual JSON Creation

Create salary_data.json with this structure:

{
  "metadata": {
    "source": "My Union Statistics 2025",
    "index_baseline": 100,
    "index_label": "Index",
    "baseline_description": "Index 100 = median salary for private sector"
  },
  "companies": [
    {
      "company": "Novo Nordisk A/S",
      "city": "Bagsværd",
      "categories": {
        "all_employees": { "count": 500, "index": 108.5 },
        "engineering":   { "count": 120, "index": 112.3 }
      }
    },
    {
      "company": "Ørsted A/S",
      "city": "Fredericia",
      "categories": {
        "all_employees": { "count": 200, "index": 105.2 }
      }
    }
  ]
}

Required fields:

• metadata.index_baseline — numeric baseline for percentage calculations
• companies[].company — full legal name (normalization handles suffixes)
• companies[].categories — object with numeric index and optional count

Option 2: Convert from Excel

Use the helper script to transform spreadsheet data:

pip install openpyxl
python3 tools/convert_salary_excel.py path/to/salary-data.xlsx \
  --source "My Salary Data 2025" \
  --baseline 100 \
  --baseline-desc "Index 100 = median salary"

The script in [tools/convert_salary_excel.py line 295](https://github.com/MadsLorentzen/ai-job-search/blob/master/tools/convert_salary_excel.py#L295) auto-detects company/city columns and outputs salary_data.json to the repository root.

Validating Your Configuration

Before using your data, verify its structure:

python3 salary_lookup.py --validate

The validation logic in collect_validation_issues() checks:

• Metadata object presence
• Companies array structure
• Required field completeness
• Numeric type consistency

Hard errors trigger fail_data_error() and abort execution; warnings print but allow continuation.

Querying Your Custom Benchmarks

CLI Usage

# Human-readable table (default)

python3 salary_lookup.py "Novo Nordisk"

# Machine-readable JSON

python3 salary_lookup.py "Ørsted" --json

How Fuzzy Matching Works

The search_company() function in [salary_lookup.py lines 63-89](https://github.com/MadsLorentzen/ai-job-search/blob/master/salary_lookup.py#L63-L89) implements:

1. Normalization: Lower-casing, stripping legal suffixes, removing non-alphanumerics via normalize() (lines 64-70)
2. Anglicization: Character substitutions via anglicize() (lines 73-78)
3. Scoring: match_score_optimized() with pre-calculated query representations
4. Filtering: Results with score ≥ 30 are retained

Programmatic Access

from salary_lookup import load_data, search_company

data = load_data()                     # validates & reads salary_data.json

matches = search_company(data, "COWI") # returns ranked list of matches

print(matches[0]["categories"]["engineering"]["index"])

Integration with Job Applications

When salary_data.json exists, the /apply command automatically:

1. Extracts company name from the job posting
2. Calls search_company() for fuzzy matching
3. Injects formatted results into the application report via format_entry()

If no match exists, the workflow continues without salary data—no blocking errors.

Key Implementation Files

File	Purpose	Critical Functions
salary_lookup.py	Core engine	load_data(), search_company(), validate_data()
tools/convert_salary_excel.py	Excel→JSON conversion	Autodetection logic at line 263
tools/README_SALARY_TOOL.md	Format documentation	Schema examples and field descriptions
tests/test_salary_lookup.py	Behavior reference	Edge cases and validation scenarios

Summary

• Place salary_data.json at the repository root to activate custom benchmarking
• Create data manually or convert from Excel using tools/convert_salary_excel.py
• Validate with python3 salary_lookup.py --validate before deployment
• Query via CLI or Python API; fuzzy matching tolerates naming variations
• Integrate automatically with /apply workflow when file is present

Frequently Asked Questions

What happens if I don't create salary_data.json?

The AI-Job-Search tool exits cleanly with an informative message, and the /apply workflow proceeds without salary comparison. No application errors occur.

How does the fuzzy matching handle company name variations?

The normalize() function strips legal suffixes like "A/S" and "Inc.", converts to lowercase, and removes special characters. The anglicize() function further standardizes international characters. Scores below 30 are filtered out automatically.

Can I use multiple data sources or update the file dynamically?

The system reads salary_data.json once per invocation. You can replace or modify the file between runs; no caching persists. To merge multiple sources, consolidate them into a single JSON file before placement.

What Excel formats does convert_salary_excel.py support?

The script uses openpyxl for .xlsx files. Column headers should include "company" and optionally "city"; remaining numeric columns become category indices. Run with --help to see all CLI options for baseline configuration.

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