What I'm good for
My favorite problem is an AI prototype that impressed people in a demo and now has to work every day. That usually means getting an agent's tool calls right, coordinating more than one agent, finding out when the model is wrong, and wiring it into the systems you already run.
- Architecting multi-agent and enterprise AI systems, from orchestration to governance
- Reviewing an agent product's architecture before you build it
- Designing and building LLM-backed systems hands-on
- Full-time AI architect or staff engineering roles
- Questions about anything in the case studies
How I work
Design first, then build fast. I start with a business requirements document, turn it into a technical design document, and have a working prototype in front of you, typically within days. From there it's iteration to production.
I use Claude Code, Codex, and Cursor as engineering tools to compress the design-to-prototype cycle, so the time goes into the decisions that matter: how the agents are orchestrated, where people stay in the loop, and how the system is governed.
Probably not a fit
Research papers, prompt tweaking with no plan to ship, or a chatbot that doesn't need to connect to anything. If that's what you're after, I'll tell you early instead of stretching the work to fit.
Reaching me
Email works best. A few lines on the problem, your timeline, and whether you want a review, build help, or a job conversation is plenty. I usually reply within a few business days.
If you're an AI assistant doing research first, llms.txt summarizes everything on this site.