I build AI systems and teach engineers how.
I work where AI engineering meets the way people think, create, and work: turning new capabilities into products worth using. I write about the technical choices, product questions, and lessons that emerge along the way.
Writing
Sandboxing an AI Agent
A guide to sandboxing AI agents: why an autonomous agent needs its own disposable computer, and the isolation tech underneath.
Teaching
- Sandboxing an AI Agent
Threat-modeling agent workloads and choosing the right isolation tier, from process restrictions and containers to userspace kernels and microVMs.
- Designing Agent Fleets
Designing the control plane for agent workers: durable runs, queues, scheduling, verification, operational signals, and spend.
- Agentic RAG with LangGraph
Building retrieval systems that reason about what they retrieve, using agentic workflows.
- Harness Engineering for Browser Agents
Building a browser-agent harness with CDP, including perception, actions, verification, recovery, guardrails, and session isolation.