I think in machines

To fall asleep I trace the refrigerant cycle through an air conditioner. Compressor, condenser, expansion valve, evaporator: the pressure and phase changes at each step, the reason the order cannot be rearranged. I do the same thing with software. I can hold a system in my head and watch it run.

That is the through-line, and it is older than any job title I have had. I built my first computer in eighth grade, a Pentium III, which dates it precisely enough.

By my junior and senior years at Roeper, a small private school in Michigan, I ran its entire Windows network. I did not help with it. I ran it. The Exchange mail server for every student, teacher, and staff member. The school website. The internal systems. Roughly 200 endpoints across the computer lab, the library, and one in every classroom: physical machines, which in 2004 meant building and deploying images far more than it meant diagnosing. I was on payroll and I had master keys to the building. I was sixteen.

Then years as a systems administrator, and a homelab that is coming up on a decade; currently forty-odd applications on a Docker Swarm running the household, managed as config-as-code.

What I am not

I am not a career software engineer, and I do not claim to be. I wrote very little production code before large language models were good enough to pair with. What I had instead was the systems picture and the SQL; I read SQL about as fluently as English, and arguably write it better.

What changed is that the gap between seeing a system and building it closed. I was early to AI-assisted development and I have stayed near the front of it, on local models and the frontier labs both. Since then: 473 merged pull requests on the platform, eighteen AWS CDK stacks operated solo, three systems built and shipped the same way.

I lead with this rather than hiding it because it is the more useful fact about me. Plenty of people can write the code. Fewer can see the whole machine, decide what it should be, and then also build it.

The work

Nine and a half years on one product. If you cannot afford groceries you go to a food pantry, which gives food away for free. Historically that meant standing in a line for hours, in public. Plentiful is how you book a time instead. I started in 2017 as the first full-time employee, when it was three pantries and 600 people, and spent that first year standing in those lines learning what the software actually had to be. It is now national infrastructure: 21 million visits, 1.3 million people, fifteen states. I was its Co-Founder and Product Director from 2022 to 2026.

The proudest part is not the growth curve. It is that during COVID, when booked appointments became the way New York City managed its emergency food lines, one site checked in 1,700 families in a single day on software designed for three pantries. It worked small, then it delivered big, without a rewrite.

Before all of it, I drove a food route for the Coalition for the Homeless: a Jesuit Volunteer Corps placement, a 50-client caseload, a commercial kitchen, a thousand meals a day. That is not a colorful origin story. It is why I can tell which decisions in this domain are real.

Now

I am in Portland, Oregon, looking for a full-time Head of Product or Director role, or a principal product or applied-AI role where building is still part of the job. Remote or Portland metro. Fractional and advisory engagements are open too, while I look.