Staff Software Engineer on the AI Platform team at SchoolAI. I build the systems AI runs inside: production inference orchestration for 1M+ classroom users, self-hosted vLLM serving, and agent control planes for long-horizon autonomous coding.
Current work
- Agent infrastructure: a multi-runtime control plane for Claude Code, Codex, and Cursor, where coding agents run 4+ hours autonomously beyond the initial task spec
- Inference: self-hosted vLLM experimentation, speculative decoding, and cost/throughput tradeoff analysis
- Training research: sleeper-agents, an SFT pipeline on a 14B distilled model documenting a negative result on date-conditional trigger learning
Background
Dual B.S. in Computer Science (Software Engineering and ML) and Mechanical Engineering (Numerical Methods and Controls), Colorado State University. Prior work spans agent platforms, ML inference R&D, and applied ML for physical systems.
Elsewhere
- Talk: "Building the Control Plane for AI Coding Agents" (ML Utah, April 2026)