Software engineer since 1979 · financial systems · multi-agent AI
I've built financial systems for forty years. The latest is an AI agent swarm that grades itself honestly.
I was wiring live market data into portfolio systems back in the 1980s — at Apple's invitation, on the original Macintosh. Today I build orchestrated multi-agent systems for systematic investing, refereed by a deterministic statistical gauntlet the agents can't game. Same problem. Forty years of better tools.
What I do
Multi-agent orchestration
Supervisor → worker hierarchies over local and cloud models, with a single control plane, cost-tiered cognition, and isolated, reproducible execution.
Evaluation you can trust
Deterministic honesty gates and a statistical gauntlet — permutation tests, deflated Sharpe, robustness and forward-decay checks — so results survive contact with reality.
End-to-end delivery
Cross-platform since the Apple II — desktop, mobile, web, cloud, Kubernetes. Data pipelines, payment/API integrations, dashboards, automation. The plumbing that turns a prototype into something that runs while you sleep.
The through-line
Forty years, one domain: making machines reason about money — and proving they're right.
Portfolio management systems in Pascal on the Apple II. The first online Dow Jones feeds, built with Apple's engineers. A single C++ source base running on Mac, Windows and Unix. Payment and API integration as the go-to specialist. A safety-tech company I founded and ran. And now an autonomous research system for systematic investing. The tools changed; the discipline didn't.
See the full heritage →Featured work
The Orchestrated Hierarchical Multi-Agent System (HMAS)
An autonomous research system for systematic investing: a coordinator drives tiers of reasoning agents to discover and stress-test strategies, while a deterministic gauntlet — not a Sharpe ratio, a conjunction of independent statistical gates — decides what's real. A swarm that doesn't get to grade its own work.
Why my systems are built to be trusted
For forty years, my other job has been keeping people alive when a wrong call is fatal.
I've spent four decades as a Mobile Intensive Care Paramedic — urban and rural advanced life support, a field training officer, a responder at the World Trade Center recovery. That work rewires how you build software. You design for the failure, not the demo. You stay calm when the system is on fire. You don't trust a number you can't verify. That's exactly why my engineering leans on honesty gates, fail-safe breakers, and evidence over optimism.
Selected writing
All writing →Have a problem that doesn't fit a template?
That's the kind I like. Whether you're a company that needs a hard system built right, or an engineer stuck on something I've already solved — let's talk.