I build systems for the part after the demo.

I build applied AI systems, retrieval pipelines, and agent architectures in Python and TypeScript. At Uncanny Owl, I help build Uncanny Automator and Uncanny Agent, working across retrieval and inference, tool orchestration, automated evaluation, observability, and the infrastructure that makes AI useful after it leaves the demo.

Models matter, but so does the system around them: the context they receive, the tools they can use, how the system recovers from failure, and the cost, latency, safety, and production behavior that determine whether an AI workflow can be trusted.

Before AI became my focus, I spent years building backend systems and commercial WordPress products. As an Envato Elite Author and Codeable Expert, I learned to ship software for real users and real constraints. That background shaped how I approach production software: clear architecture, durable state, and a bias toward systems that keep working when conditions are not ideal.

Model the domain. Make the paths boring. Match the weight of the solution to the weight of the problem.