A selection of projects, chosen because the engineering problems were interesting rather than because they're the most recent. Each one is written up in full — including what was built, what wasn't, and what I'd do differently.
ViroTrade: proving a renewable energy certificate is real
Independent audit of renewable energy certificates, where proof arrives as whatever a registry happens to emit: a CSV, a PDF, an email, a screenshot. I built the platform that ingests any of those, extracts the transaction data with AI, and compares it against data pulled directly from the registries — two independent paths, so a fault in one can't manufacture agreement with the other.
Sole engineer, architecture through deployment. Java 21, Spring Boot, React, MySQL, and a multi-provider AI layer spanning OpenAI, Anthropic, Google, and OpenRouter.
Greenman: trading 2,000 symbols with nobody watching
A day trader wanted his strategy to run without him. The system tracked roughly 2,000 symbols across four intraday timeframes, evaluated his strategy against all of them continuously, and traded through Interactive Brokers — unattended, every trading day.
Built in 2001, before any of the infrastructure this needs existed. The hard part wasn't the strategy logic; it was operating correctly on top of market data feeds and broker connections that failed constantly, with nobody watching.
Quokka: an AI companion older adults would actually keep using
A relationship-first AI companion for older adults, co-founded with a former Nissan marketing executive. I built the Flutter client and the Spring Boot backend: offline-first, with every write hitting the local database before the network, and accessibility stored as a property of the person rather than the device.
The interesting decision was what I deliberately left out of the phone. The product is longitudinal, which means it needs memory — and memory that has to survive device loss and serve two different customers can't live in the client.
CPSS: a small app built like a real system
A salad builder that scores combinations across crunch, punch, sweet, and savory while computing nutrition live from per-100g ingredient data. A small, fun project, built with the same architecture I'd use for client work: layered separation, domain models kept distinct from persistence, real auth, versioned migrations.
Also
Twenty-five years of prior work at Cisco, Amelia AI, Dell, Disney, Tokyo Electron, and others — reporting infrastructure, NLP and automation, FinTech, and AI-powered recruiting platforms.