I work on agents at Databricks — harnesses, evaluations, infrastructure, memory, and the systems that tie them together. I'm especially interested in long-horizon agents, agent harnesses, and post-training models that improve agent quality and performance.
Before this I did an M.S. in CS at Princeton with the NLP group and a B.A. in CS at Cornell, with ML research stints at Scale AI and Snap. Outside of work I cycle, hike, run, swim, and eat.
Things I've worked on
- Omnigent — open-source meta-harness for Claude Code, Codex, and Cursor
- Databricks Supervisor Agent — managed orchestrator for Knowledge Assistants, Genie Agents, and tools
- Supervisor API — programmatic API that runs the agent loop for custom Databricks agents
- Databricks serverless forecasting — AutoML time-series forecasting on fully managed compute
Side quests
- What's growing here? — a playful map of what California grows, from DWR field-level crop data Data: California Department of Water Resources, Statewide Crop Mapping 2023 (Land IQ) · © OpenStreetMap contributors, Protomaps · AWS Terrain Tiles · Microsoft Fluent Emoji (MIT) · Figtree (SIL OFL 1.1)