Those questions follow me around. Whether I’m looking at a machine, a building, or how people work together, I tend to work backwards from what I can see to understand how it got that way.
That curiosity runs through my work, from aviation and fabrication to software and robotics. I like understanding how the parts fit together, what they’re meant to accomplish, and what happens when they don’t.
Some projects are released, others are still taking shape. Here’s what I’ve been working on.
Air traffic simulation with an AI agent
Built in about five days for the WebMCP Challenge on Devpost. A human and an agent share a live air-traffic simulation, using tools defined by the application and acting within the authority the person grants.
Local models for the pi coding agent
An extension connecting the pi coding agent to local models through Ollama. I maintain it, work through user reports and help contributors improve their PRs, with the aim of building a community around the project.
Certification study for iPhone and iPad
A study app for iPhone and iPad, built to learn the process of taking an idea through to App Store approval. I kept the scope simple and chose something I might use for my own cybersecurity studies. Released on the App Store.
Hardware, navigation and agent interaction
An ongoing robotics project bringing together hardware, navigation and agent interaction. I’m exploring how a robot can move through a space, make sense of its surroundings, and interact with people within clear limits.
I use coding agents throughout development. The useful work starts with making the problem clear, choosing where responsibilities belong, and giving each piece a way to be checked. I work through the implementation with the agent and compare the result with what I actually intended.
A passing test can still miss the thing that matters. On the robot, a correction that worked in synthetic rooms made navigation worse in the real one. I removed it and recorded why. Those decisions stay with the project, so a later session can build on the lesson.
Long projects outlast a conversation. I developed a briefing workflow that keeps current state separate from lasting guidance, so the next session can pick up the work. Related projects feed into a shared overview.
I’m primarily self-taught, with a background in commercial aviation, running a makerspace and hands-on fabrication. I’ve spent a lot of time learning how unfamiliar equipment works and what it takes to make it useful.
That carries into the software and robotics here. I enjoy the details, and I like working with people who bring a different set of questions.