I've spent twenty years making technical systems people can trust. AI is just the newest one.

It started in the field.

For two decades I designed and troubleshot industrial moisture-control systems -- the unglamorous kind, where humidity is a process variable and being wrong shows up as spoiled product, a failed batch, or a plant that has to stop. I came up the hard way: hands-on troubleshooting first, then design and applications engineering, then sales and operations leadership, across well over a hundred facilities in North America, Europe, and Asia. That work taught me the thing that turned out to matter most later -- a system you cannot measure and cannot trust is not a system. It is a liability with a power cord.

Then I started building.

I co-founded Desiccant Air Solutions, an engineering company built around a smarter approach to industrial dehumidification, and in 2022 I started Harvey Consultancy to do independent work for clients across the US, Europe, and Asia. Along the way I built the tools the work needed -- calculators, sizing programs, custom software -- because the fastest way to make hard technical work repeatable is to build the thing that does it. I am an electrical engineer, an MBA, and a PMP, which is a long way of saying I think in systems and I run things like projects.

Then I went all the way down the rabbit hole.

Not the headlines -- the engineering. I wanted to know how these systems actually get built, where they create real value, and where they quietly fall apart. So I built. A production-grade research system that answers questions from a large, messy body of source material with citations back to the exact source -- built to a standard where a wrong answer that cannot be traced to its source is a failure. A multi-agent pipeline that writes technical documents and checks its own work. Research assistants, a phone agent, monitoring agents. The standard I learned in the field carried straight over: an answer you cannot trace is one you cannot use.

Scoped, plain-spoken, no lock-in.

I run things the way I learned to run projects -- one step at a time, with a clear deliverable at each, and nothing that locks anyone in. I would sooner say "AI is not the answer here" than sell something that is not. That is about the whole of it.

When I am not building something, I am usually around Buffalo and Grand Island, training for the next race.