US Job Market Visualizer: Explore BLS Data and AI Exposure Across 143M Jobs

⬅️ Back to Tools

TLDR

  • US Job Market Visualizer is an interactive treemap of 342 occupations from the Bureau of Labor Statistics, covering 143M jobs across the US economy.
  • Area represents total employment; color represents the metric you select.
  • Toggle between BLS projected growth, median pay, education requirements, and Digital AI Exposure.
  • The AI Exposure layer is powered by an LLM prompt pipeline: write a prompt, the model scores each occupation, and the treemap recolors.
  • Open source at github.com/karpathy/jobs. You can run custom prompts for any criteria you want.

What It Is

karpathy.ai/jobs is a research tool by Andrej Karpathy that turns dry Bureau of Labor Statistics tables into an explorable visual interface. Each rectangle in the treemap is an occupation. Bigger rectangles mean more people work in that job. Color intensity shows how that occupation scores on your selected metric.

This is not a formal economic report. It is a development tool for exploring BLS data visually and testing how LLMs reason about real-world occupational descriptions at scale.

How the Visualizer Works

The Treemap

  • 342 occupations from the BLS Occupational Outlook Handbook
  • 143 million total jobs represented
  • Rectangle area = total employment in that occupation
  • Rectangle color = score on the selected metric

Click any tile to open the full BLS page for that occupation.

Built-In Layers

LayerWhat It Shows
BLS OutlookProjected growth or decline from the Bureau of Labor Statistics
Median PayAnnual median wage for the occupation
EducationTypical entry-level education required
Digital AI ExposureLLM-estimated score of how much AI will reshape the job

The AI Exposure Scoring System

The Digital AI Exposure layer uses a custom LLM prompt that scores each occupation from 0 to 10. The model reads the official BLS description and rates how exposed the job is to AI-driven transformation.

Key signals the prompt uses:

  • Digital vs. physical work: Jobs done entirely on computers score higher (7+) because AI advances fastest in digital domains
  • Direct automation: Can AI do the core tasks today or soon?
  • Indirect effects: Will AI make workers so productive that fewer are needed?

The scoring anchors:

  • 0-1: Minimal exposure (roofer, landscaper, commercial diver)
  • 2-3: Low exposure (electrician, plumber, firefighter)
  • 4-5: Moderate exposure (nurse, police officer, veterinarian)
  • 6-7: High exposure (teacher, manager, accountant, journalist)
  • 8-9: Very high exposure (software developer, designer, translator, data analyst)
  • 10: Maximum exposure (data entry clerk, telemarketer)

Important caveat: These are rough LLM estimates, not rigorous predictions. A high score means the job will likely be reshaped, not necessarily eliminated. Software developers score 9/10 because AI is transforming their work – but demand for software could grow as each developer becomes more productive.

Why It Matters

  • Instant macro perspective: See the entire US labor market as a single visual field
  • Compare across dimensions: Switch from pay to growth to AI exposure in one click
  • LLM as analyst: The tool demonstrates how LLMs can be used to score and color real-world datasets by any criteria you define
  • Customizable: Fork the repo and write your own prompts for robotics exposure, offshoring risk, climate impact, or any other question

How to Run Your Own Prompts

The source code includes scrapers, parsers, and a pipeline for running custom LLM prompts over all occupations. To create your own coloring layer:

  1. Clone the repository:

    git clone https://github.com/karpathy/jobs.git
cd jobs

  2. Write a scoring prompt that asks the LLM to rate each occupation on your chosen dimension

  3. Run the pipeline to score all 342 occupations

  4. The treemap recolors based on your custom scores

This makes the tool a general-purpose framework for LLM-powered occupational analysis, not just a single visualization.

Tips

  • Start with Total jobs view to see which occupations employ the most people
  • Switch to Digital AI Exposure to see which large-employment sectors face the most transformation
  • Click individual tiles to read the full BLS description and verify whether the LLM score makes sense
  • Remember that high AI exposure does not mean job elimination; many high-exposure roles will be reshaped, not replaced
  • If you fork the project, try scoring by “remote-work feasibility” or “humanoid robotics exposure” for alternative perspectives

That is it. A single visual interface that turns 143 million jobs into an explorable map, with an LLM pipeline that lets you ask your own questions about the future of work.

Crepi il lupo! 🐺