I am a principal GenAI engineer. I build production AI agents: systems that manage product backlogs, review code, and triage production issues, plus the platform and orchestration layer underneath them. Most of my writing here is about the parts of enterprise AI that decide whether any of it works: process, ownership, evals, and what happens when things fail.
What I work on
The short version of the last few years:
- Backlog management agent. Persona-based agents that run our scrum workflow, triage, assignment, and follow-ups, tuned to how our teams actually work.
- Code review agent. Tiered review across a large pull request volume, with cost treated as a design input from the start rather than a surprise at the end of the month.
- Production triage agent. First-pass diagnosis on production issues, so an engineer starts from a hypothesis instead of a blank page.
- Agent orchestrator. One control plane where agents are discovered, routed, and governed, instead of every team wiring its own point-to-point version.
- Model routing. Routing decisions driven by tool-call failures, task complexity, and sub-agent behavior, with escalation when confidence drops.
- Platform plumbing. MCP servers and an enterprise deep wiki, the unglamorous layer that lets agents reach the tools and the knowledge they need.
What I think about
The essays on this site circle a few questions I keep coming back to. When should you build an agent and when should you buy one? How do agents earn trust: identity, evals, accountability when the output is wrong? What changes when systems stop waiting to be asked and start acting on schedules? And what happens when the binary itself becomes an attack surface? If you read one piece, start with the flagship essay.
Beyond work
Metrocrest Toastmasters on Tuesdays, because presenting is a skill like any other and it improves with reps. The DFW AI community. Board games, video games, photography, and travel. I live in Flower Mound, Texas, with my family.
Elsewhere
LinkedIn is the best way to reach me. I read everything, and I reply to most of it.