13 years
End-to-end delivery since 2013: the conversation, the design, the build, the handoff
Meridian, Idaho · Designer · Developer · Founder · Fractional Chief AI Officer
I sit with a business, figure out how it actually works, and build what it needs.
Thirteen years of that: client conversation, design, build, handoff, for Hawaii businesses, Silicon Valley software companies, and two companies of my own. Since November 2025 the build is AI: systems I design, run on my own hardware, and answer for myself. Talk to a human who builds AI.
Track record
2014 → 2024
Andersen Digital, my first consulting client. Still collaborating a decade later, most recently on AppOmni
Rehired
AppZen brought me back in Nov 2022 after two years leading their site. Clients rehire; that is the track record
213 runs
Production agent runs I measured before trusting my own fleet: 46% finished at baseline, so I built the checks
Featured work
Prismatic Engine
Open-source auditable validation hypervisor (AGPLv3). It sits over my agent fleet on any major OS or environment, and I check every output before it counts.
Human Design Engine
A commercial chart-computation API and MCP server. A product I built, priced, and run.
BeyondSaaS.ai
I help companies put AI to work inside their own operation, starting with a paid review of where it actually pays.
Production AI Infrastructure — Self-Hosted
Most consultants rent their AI. I run mine: a Proxmox cluster and GPU pool in my own rack, serving everything on this site.
agent-checkpoint
A published open-source plugin (MIT, v0.3.0) that makes agent runs leave a checkpoint on the record, so a run that hits its budget still leaves its work behind.
Sentinel IT Asset Disposal & Logistics
Secure IT asset disposal in the Treasure Valley: scheduled pickups, documented chain of custody, and a business that runs on tooling I built. It keeps my hands in hardware and operations.
How I work
Start with the people who do the work
I sit with them and map how the process really flows, not how the org chart says it flows. Effort, risk, and handoffs included. I have done this since 2013; the discovery phase is the job.
Cut before you automate
Steps that add no value come out first. Automating a broken process just makes it fail faster.
Build it, by hand first
Prototype small, with the team using it. Agents take the steps they can prove they handle, with tool access and permissions scoped to each step, nothing wider.
Keep a human on the controls
Approval checkpoints where judgment matters. In my own operation, payments, deletions, and money movement stay human. That line is deliberate, and it is mine to draw in any system I build.
Measure it, write it down
Baseline first, then check at 30 and 90 days. What worked becomes a playbook the team keeps. When I pointed this at my own fleet, the baseline said 46%, so the fleet got better checks, not better excuses.
The short version
I started in Hawaii in 2013: graphic design degree, then straight into client work: radio stations, a health insurer, an enterprise security company's website for five years. I co-founded a kayak tour company the same year I graduated and still run it, remotely, from Idaho. Silicon Valley clients followed. AppZen made me lead developer on their site for two years, then rehired me. Since November 2025 I work full-time on AI: a fractional CAIO practice, an open-source engine, and the hardware it all runs on. Different tools every era. Same job: figure out what a business needs, build it, and stay for what it changes.
Current work: BeyondSaaS.ai ·Prismatic Engine ·agent-checkpoint ·Human Design Engine ·github.com/mbgulden
More about meBuilding AI into your operation?
I take on a small number of fractional CAIO and AI implementation engagements. Email is the fastest way to reach me.