US Job Market Visualizer – Andrej Karpathy
Interactive tool lets you write LLM prompts to score any job's risk from AI, robotics, or climate change.
Former OpenAI researcher and AI educator Andrej Karpathy has released an open-source research tool called the US Job Market Visualizer. Hosted on GitHub, it creates an interactive treemap of the entire US labor market, visualizing 342 occupations from the Bureau of Labor Statistics (BLS) that represent 143 million jobs. Each rectangle's size corresponds to total employment, and users can toggle views to show metrics like median pay, education requirements, or projected growth. The core innovation is its modular, LLM-powered scoring pipeline.
Instead of providing a single, fixed analysis, Karpathy built the tool for exploration. The source code includes scrapers and parsers for BLS data, connected to a system where anyone can write a custom Large Language Model prompt to score occupations by any criteria. The included example prompt scores 'Digital AI Exposure' on a 0-10 scale, assessing how much current AI could reshape a job based on whether its work product is digital. Users can modify this prompt to analyze exposure to humanoid robotics, offshoring risk, or climate impact, generating a new visual coloring instantly. Karpathy emphasizes these are rough LLM estimates, not predictions of job loss, as high exposure often means transformation, not replacement.
- Visualizes 143 million US jobs across 342 BLS occupations in an interactive, size-scaled treemap.
- Features a unique LLM pipeline: write a custom prompt to score and color jobs by any user-defined risk (AI, robotics, climate).
- Open-source on GitHub, turning static government data into a dynamic tool for economists, policymakers, and researchers.
Why It Matters
Provides a flexible, data-driven framework for professionals to model and visualize the future impact of technology on specific careers and industries.