production systems shipped
unit property portfolio deployed
verticals with shipped work
operator end-to-end
Three things every demo skips.
Launch-ready, not draft.
Every deliverable ships complete: ad copy with image direction, landing pages numbered and ordered, email sequences with send timing, lead forms with qualification logic. Not skeletons. Not placeholders. The version you see is the version that goes live.
A separate agent audits every output.
Producer agents work in parallel. A reviewer agent — separate from them — runs voice consistency checks, numerical conflict scans, and compliance audits on everything they made. Recent project: 4 compliance findings flagged, 12+ numerical conflicts surfaced, all caught before delivery.
Multiple models. Cross-checked outputs.
Outputs route through multiple AI providers — different models for reasoning, multimodal, and long-context — and verify against each other. When one drifts on tone or hallucinates a fact, another catches it before the output ships. The difference between a chatbot that sometimes works and infrastructure that runs.
Anyone can run ads.
The moat is underneath.
From a recent build — Custom home building / Treasure Valley
Running in production right now.
- StereotaxisClinical Specialist, robotic cardiac ablation, North America
- TARUFounder, AI recovery platform, two incubators, regional $1M VC competition
- CorComProject Manager, 15 concurrent research engagements, 15+ client POCs
- Ko Awatea (NZ)RCT design under published suicide researcher, Outstanding Original Research award
The work is getting skilled people to trust an unfamiliar machine in a high-stakes moment.
For eighteen months at Stereotaxis, I owned physician adoption of robotic cardiac ablation across North America — getting thirty-year cardiologists to trust a machine with a catheter inside a beating heart. The robot worked. Adoption was the actual job. Same pattern at TARU, the AI recovery platform I built and shipped inside clinical settings: the model was the easy part. Getting a therapist to route a patient to it was the real lift. Larkin Systems is that pattern productized for mid-market operations — brokerages, M&A firms, boutique hotels, enterprise cleaning. Same engagement shape every time: ninety days, one bottleneck, a working system your team actually uses by week six.
“Understands science, complexity, systems thinking, and non-linear creative problem-solving.”
What people say after the work ships.
Luke Larkin helped me get more out of AI in my coaching and consulting business, and he has a rare gift for making the technology practical and immediately useful. I came in wanting to understand AI agents, Claude projects, and how to use these tools for content, outreach, and execution. Luke quickly zeroed in on the real issue: I had a lot of valuable material sitting inside Claude, but I hadn’t given the system enough structure to actually put it to work. He helped me rebuild a Claude project into something that functions like a real AI agent — clear purpose, focused files, strong instructions, a defined workflow. That was the turning point. Rather than chasing abstract AI ideas, he showed me how to sharpen the tools I was already using. What stood out was the combination of technical fluency and business judgment. He knows the tools, but more importantly he knows how to apply them to real workflows without overcomplicating things. If you’re a founder, operator, or consultant trying to move from experimenting with AI to actually running it in your work, Luke is who I’d recommend.
I’ve worked with a lot of technical folks at MIT and in my years scaling startups, so I’m selective about who I partner with. Luke’s ability to develop custom Claude skills, work in multi-model systems, and evolve prototypes into robust production deployments is a wildly useful combination. When something needs to ship, Luke is the perfectionist I want in the room.
If you’ve got a system that should be running and isn’t, book the call.
No pitch deck. No proposal template. Just the call.