Founder-led AI consulting
I help teams move AI into the places where mistakes cost money, speed matters, and people still need to trust the result.
I spent more than 20 years building products and companies. I founded Browntape, scaled it into a serious commerce operations platform, and led it through acquisition. Now I work with a small number of clients who want AI embedded into real workflows, not pasted on top for a presentation.
Direct founder access. A small number of consulting engagements at a time. Built for implementation, not theatre.
The Truth
They fail in the handoff. In the review loop. In the missing audit trail. In the part where nobody decided what happens when the model is uncertain and a human has to step in.
That is the part I care about. I design systems where AI can do useful work inside an actual business process, with enough structure that a team can keep using it after the novelty wears off.
I am not interested in generic AI transformation language. I am interested in whether the workflow gets faster, the quality gets better, and the team can rely on it next month.
Not a prompt pack. Not a strategy deck. A working system, with edges, controls, and consequences understood in advance.
What You Won't Get
What I Will Build
The model matters. The orchestration matters. But the workflow matters first. That is where the leverage is.
Selected Work
I built an agentic lesson-plan system for structured curriculum production. This was not a loose content-generation experiment. It was a six-step workflow with gating, reusable visual assets, translations, student materials, and a review loop serious enough for editorial use.
The value was not just that AI could write. The value was that educators and editors could work inside a system that made outputs reviewable, versioned, and reversible.
Wrong measurements create returns. Returns destroy margin. I worked on an AI-assisted furniture measurement workflow that compared browser photogrammetry, cloud reconstruction, and LiDAR-based capture paths to find something customers could actually use without a complex setup.
This is the kind of project I like: real constraints, expensive mistakes, and a technical path that has to make sense in both UX and business terms.
I built a two-stage portrait pipeline that turned a visitor photo into a Mario Miranda-style illustration using sketch generation and custom LoRA fine-tuning. It had to feel magical, yes. But it also had to be reliable enough to run live for the public.
That meant model selection, training-data correction, bias handling, and an experience design that could hold up beyond a single demo day.
Browntape orders generate deep lifecycle logs. I built an MCP-based workflow that let AI agents inspect structured Elasticsearch log data and draft root cause summaries before a support person even opened the ticket.
That is AI diffusion in the form I respect most: less manual triage, clearer context, and a better starting point for the team already doing the work.
Why Clients Call Me
I have built products from scratch. I have scaled systems under real load. I have hired teams, shipped through ambiguity, dealt with customer pain, and carried the responsibility when the system had to keep working.
So when I look at an AI opportunity, I do not just see a model choice. I see support burden, edge cases, adoption friction, and whether the economics still make sense six months later.
Contact
Send me the process that feels expensive, slow, brittle, or too dependent on human effort. I will tell you quickly whether AI belongs there, and if it does, what shape the first serious version should take.