The background behind the practice

I've spent over twenty-five years building and operating production software at Amazon, OJO Labs, Rackspace, and Bazaarvoice, through three IPOs and eight acquisitions. My work is the operational layer around the code: the pipelines and platforms that let engineering teams ship reliably at scale. Now I apply that experience to production AI agents.

AI accelerates how fast code gets written, but the review, security, cost control, and observability around it do not speed up on their own, and that gap is where shipping fast turns into runaway cost and outages.

I build on the standards the field is converging on: evolving principles for structuring agentic systems, the Cloud Security Alliance's Agentic AI Runtime Mitigations (AARM) specification for runtime security, the OpenTelemetry GenAI semantic conventions for observability, and the Model Context Protocol for tool and data access. That keeps the work portable across harnesses and legible to the platform teams who inherit it.

I work with a small number of clients at a time so I can stay close to the actual problem.