Senior PM→Solo AI Founder
A decade turning complex systems into things people can actually use — inside UBS, Credit Suisse, and high-growth startups. Now a one-person AI studio building autonomous agents and the interfaces you talk to them through.
I’ve spent my career turning complex systems into things people can actually use — first inside global banks and fast-scaling startups, where the real job was making intimidating software feel obvious. The pivot to AI wasn’t a break from that; it’s the same instinct with sharper tools. Autonomous agents are the most capable systems I’ve ever worked with and some of the least usable — so I build both halves: the agents themselves, and the interfaces that make them feel less like a command line and more like something you can talk to.
I run AMPMIO as a studio of one that operates like a small team — because I think that’s what the future of work looks like: people orchestrating fleets of agents, with judgment and taste as the human contribution. I’d rather build in that future than predict it. I’m drawn to the engineering and the philosophy in equal measure: what these systems can do, and what it means that we now talk to them as if someone’s there. This site is that thesis in miniature — a terminal that renders in real time and answers questions about the work.
Production multi-agent AI systems and the operations dashboards that run them — orchestrating fleets of coding and ops agents, with the trust boundaries and human-in-the-loop controls that make them safe to actually deploy.
A browser-based multiplayer traversal game set in Neukölln-9, a near-future cyberpunk Berlin. Cel-shaded 3D, real-time multiplayer — a proving ground for interactive-web craft: physics, netcode, and feel. Public demo in 2026.
High-craft, diegetic web interfaces — this site included. WebGL, shaders, and motion in service of products that feel like an experience, not a template.
An always-on personal AI agent: memory, tool use, self-hosted infrastructure, reachable from my pocket. The roadmap points at hardware — giving it a body is the next experiment.
A generative interface system for AI agents — UI that composes itself to fit the answer, instead of forcing answers into fixed screens.
A multi-agent simulation where AI agents develop culture, trade, trust, and conflict on a shared grid — an open experiment in what emerges from a society of agents.
A retro arcade roguelike — whack-a-mole run through a run-based scoring engine. In development.
$open channel — for roles, collaborations, or a look under the hood.
draft · links pending confirmation