Computer Science and Mathematics at the University of Kansas. I build agentic LLM systems people can check, and I'm working toward frontier AI that's both capable and interpretable.
- NASA/JPL, 2026: Built an agentic LLM system for technical analysis and decision support.
- Location memory: Self-hosted MCP servers connecting personal location data with LLM tools.
- Canvas MCP: Extended an open-source Canvas LMS MCP server with additional browser and file capabilities.
Mechanistic interpretability · Representation learning · Generalization · AI agents