Enterprise Technology Executive · Greater Milwaukee
For 20 years I’ve led IT strategy, budgets, and transformation at global manufacturers like Rockwell Automation and Generac — aligning technology with what the business actually needs. I also build and run production AI systems myself, which keeps my judgment honest: I can tell a demo from something that will survive in production and earn its cost. The work below is where I prove it.
I’ve spent 20 years leading enterprise technology — the strategy, the budgets, and the teams that deliver it — at companies like Rockwell Automation and Generac.
My job is turning technology into business results: aligning what we build and buy to what the business actually needs, governing the investment, and getting it into production. Along the way I’ve directed a $336M budget, taken roughly $16M a year back out in cost, integrated acquisitions, and moved AI from pilot into governed production.
What makes my read on AI different is that I don’t just advise on it — I build and run it myself. The systems below are production AI I operate on my own hardware. They’re how I learn, firsthand, where autonomous AI is genuinely ready to own real work and where it still breaks down.
That keeps me honest when it counts. When I tell a leadership team what AI will and won’t do for them, it’s grounded in what I’ve actually shipped and watched run in the real world — not a vendor demo.
How I build AI that holds up in production.
Agent Runtime
A persistent orchestration engine for autonomous AI. Loom coordinates multiple agents, preserves long-term memory, and continuously rebuilds context so work can continue across interruptions, model limits, and restarts. Rather than treating each interaction as isolated, it gives AI systems durable state, recoverability, and the ability to pursue long-running objectives with minimal human intervention.
Loom is no longer just a design. It runs persistent agents in production today — coordinating real, long-running work without losing the thread across restarts and model limits. The runtime stopped being a concept and became infrastructure I depend on daily, which is the only test of an architecture that matters.
Because autonomous systems shouldn’t forget who they are.
Read how it works — the architectureAgentic research platform · concluded
ATLAS was a research platform for agentic AI disguised as a trading system. Inside a tightly constrained portfolio, it researched, planned, executed, and reviewed its own decisions, so I could watch how an agent handles reasoning, risk, and feedback loops with real money on the line. I wrapped the experiment when fully autonomous, no-human-in-the-loop trading was no longer permitted. That ending was the clearest result of all: these systems belong with a person in the loop, not with the person taken out of it.
A finished experiment. The lesson that stuck: keep a human in the loop.
Autonomous voice agent
An experiment in real-world AI agency. Concierge autonomously places calls, navigates conversations, discloses its AI identity, completes routine tasks, and reports back with structured results. The project explores the practical challenges of deploying trustworthy voice agents in environments designed for humans.
It handles the conversation. I handle the exceptions.
Self-improving quality layer
Running autonomous agents at scale, the hard part isn’t any single agent — it’s keeping them honest. Every agent checks its own work against a pass/fail number before it hands anything off, so it corrects against data instead of vibes. When a job fails, it routes itself to an agent that diagnoses it, fixes it, and reports back what it did — no human needed.
Once a week the whole fleet grades itself on yield, cost, and drift, and proposes its own improvements. I approve the ones worth building. The machine handles detection and the routine repair on its own; I stay in the loop for the calls that need judgment. That balance — systems that run themselves, but know when to ask — is the whole point.
It catches its own mistakes. I make the calls that matter.
Real systems, with real users and outcomes.
Government stock-trade tracker
A nonpartisan intelligence platform that transforms government financial disclosures into structured, searchable data. Covering Congress and senior executive-branch officials, it delivers weekly insights and exposes an MCP API that enables AI systems to reason over government trading activity.
potomacledger.comTechnology studio & advisory
Seven Roots is an independent technology studio focused on building reliable AI systems that do useful work. Every project begins as an experiment in autonomy and ends only when it can be trusted to operate in the real world. The studio is still evolving, and I’m always interested in conversations with people who care about thoughtful engineering, practical AI, and long-term impact.
sevenrootsconsulting.comAI technology-decision advisor
An AI system for high-stakes technology decisions, producing structured, transparent, and defensible reasoning for complex platform and architecture choices.
heartwood.sevenrootsconsulting.comFertility-navigation platformSupporting · Coming soon
A technology-enabled fertility navigation platform designed to make one of healthcare’s most complex journeys more accessible and easier to navigate. By combining frontline nursing expertise with trustworthy AI, it helps patients better understand treatment options, improve medication access, and make informed decisions with confidence. Founded by my wife, an experienced nurse, I lead the technology, AI architecture, and product development to build intelligent systems that responsibly augment—not replace—clinical guidance.
Smaller builds — testing ideas on real, everyday problems.
Autonomous events newsletter
A fully autonomous local discovery agent for busy families. Every week it searches southeastern Wisconsin for the best kid-friendly events, filters the noise, writes a curated newsletter, and sends it on schedule—all without anyone pressing “send.” What began as a simple automation became an exploration of how AI can quietly handle recurring, real-world tasks people genuinely value.
Every Sunday. Without me.
Self-maintaining media catalog
An autonomous media catalog for the AI-generated music I create. Every new track triggers an agent that ingests the audio, extracts metadata, generates artwork and a song summary, and organizes everything into a searchable, self-maintaining library. Built to explore how AI can continuously structure and manage growing collections without manual curation.
It organizes itself.
Local-government meeting monitor
Local government decides a lot in public meetings almost nobody reads. Docket watches the agendas and minutes across a whole metro, pulls out what’s actually being decided, and flags the businesses it affects before the vote lands. The newest experiment here — still taking shape.
Reads every room, so you don’t have to.
Autonomous two-seat cinema
A headless system that records one movie a week off the antenna that fits both my wife’s taste and mine. It reads the over-the-air guide, scores every candidate against two taste profiles, picks the one that clears a floor for each of us, sets the recording, and sends the pick ahead of the airing with a one-tap swap. After movie night, a thumbs-up or thumbs-down quietly tunes our profiles for next time.
Two people, one pick, no negotiation — a small, human problem solved by an agent that decides, acts, and learns on its own.
One movie a week. Chosen for two, picked by itself.
If you’re trying to tell what AI will really do for your business from what just demos well, that’s a conversation I enjoy. Same if you want a technology leader who has done both.
Let’s talk
