Solo full-stack builder · AI/ML · simulation
Seventeen years making 3D simulators taught me how real software actually gets shipped. These days I take an idea — mine, a friend's, a client's — and turn it into a working product. Fast when speed is the point. Sound enough to keep when it isn't.
The one thing worth knowing
I have an almost unreasonable allergy to that word — because most of the time "impossible" isn't a verdict, it's a shrug. It means I don't want to, or I haven't bothered to think it through. And I've never been able to take that as an answer. Tell me something can't be done and I'll go find out for myself — and more often than not, it turns out it just hadn't been tried.
What this actually means
Sometimes you need something real in front of users or investors by next week — a prototype that isn't a throwaway. Other times you've got a genuinely hard system that needs one person who can hold the whole thing in their head — physics, ML, backend, cloud — and make the calls that keep it coherent. The two projects below are one of each.
The receipts — working software, not slides
Fast — the weekend build
Idea to a live product in a weekend. A word-game concept, taken from nothing to a tested game engine, a brand, a marketing site, and self-deploying infrastructure — about 36 hours of work across two days. The game balance was tuned against a Monte-Carlo simulator, not guesswork.
Hard — the deep build
A six-domain AI platform, built solo. A physics-accurate locomotive simulator, a reinforcement-learning pipeline that trains an AI to drive against that exact physics, a C++ desktop app, and a full multi-tenant B2B platform on AWS — roughly 116,000 lines, architected and built end to end by one person. The kind of thing most teams split across specialists.
The long way here
I started as a 3D artist — modeling, texturing, lighting, the craft of making something look like it belongs. Then I spent seventeen years at a simulation company, where I went from building the art to leading production on more than fifty large, technical projects, wrangling teams of artists, developers, and specialists through the whole messy lifecycle.
The catalyst was a bug. For a long stretch I had to live with a severe defect in the software every single day — and every time I raised it, the answer was the same: impossible. A feature I wanted? Impossible. An idea I had? Impossible, for one invented reason or another. After a while it was clear "impossible" didn't mean can't — it meant won't: nobody wanted to do the work, or sit down and reason through it. Being expected to just accept that was the part I couldn't stomach.
So I taught myself to program — Python, then C++, then whatever the problem demanded — and fixed it myself.
The thing I'm chasing is the same now as it was at the start: the moment a half-formed idea suddenly runs on a screen. Now I build entire products on my own, using modern AI tooling to move at a pace that used to take a whole team. The art training never left, either — I can make a thing work and make it feel like something, which turns out to be a rarer combination than it should be. And I still hear "that can't be done" as a starting line, not a stop sign.
The kinds of things I make
From the idea on a napkin to a deployed, tested thing people can actually use — frontend, backend, and the deploy pipeline that ships it.
Physics engines, reinforcement-learning pipelines, and training/evaluation systems built so the numbers you trust match how the thing behaves when it ships.
AWS, infrastructure-as-code, CI/CD that deploys itself with no secrets lying around. The boring, reliable parts, done right.
I came up as a 3D artist, so I can shape how a product looks and feels — brand, UI, real-time 3D — not just whether it compiles.
For the people who scan for keywords: C++, Rust, Python, TypeScript/React, AWS (Terraform, Lambda, Cognito), Docker, reinforcement learning (PPO, Stable-Baselines3, ONNX), MCP servers & agentic workflows, and a full 3D toolchain (Maya, Blender) when a project calls for it.
Is this a fit?
You know what you want the system to do. You need someone to actually make it do it.
Multi-domain, a little scary, the kind of thing that needs one head holding all of it.
Something real for users, a pitch, or validation, built well enough to keep.
No form, no funnel. It goes straight to me.