How I work

Most of my work starts before any code, with reframing the goal. “Add AI” isn’t a plan. So I sit with how someone actually spends their day, find the one task that eats their time, and turn a vague wish into a concrete, measurable scope. Then I build the smallest thing that wins that time back. With vibe coding I can usually get a working version in front of them in a day or two, so they feel the difference before committing to more. It tends to come out two ways:

• Enable. Get a team using AI tools on their own, so the skill stays with them after I leave.
• Build. When the workflow needs more than a tool, I build the AI service for it.

Why now

I started as an AI researcher, became a CTO, and now run a small practice where I speak the client’s language. For a long time those felt like separate chapters. Lately they’ve become one. The researcher in me can build the hard part. The CTO can talk across very different engineering teams. The pragmatist can read a business and the people in it. I was always more interested in technology being used than in technology for its own sake, and the recent jump in AI is what finally let me use all three at once.

Where you can see it

Pragmatist is the day job now, a small AI-consulting practice for first-time adopters. Seven projects across e-commerce, healthcare, edtech, and media, every client coming back. The pattern repeats: take a daily manual chore and turn it into a job that runs itself. Four hours of price monitoring becomes a 30-minute batch. A full day of paperwork becomes one click.

ARKHE Animation is a four-person team making a Bible-themed animated series with generative AI. One of us is a pastor’s wife who had never really used a computer. I taught her our tools, and three years on she runs the production herself with Claude Cowork. That’s the kind of outcome I care about. We work with Midjourney, Runway, Kling, VEO, Nano Banana, ElevenLabs, SUNO and others, and this year I wired the team’s tools into one agentic workflow (Claude and MCP) so the non-technical members can run multi-step jobs without waiting on me.

The deep builds are where I go when a problem needs research depth. LenaLab is the clearest case: my attempt to fuse the research background I grew up in with the recent leap in AI agents, an autonomous lab where an AI agent writes and verifies its own computer-vision algorithms. Before that, real-time 3D tennis tracking deployed at facilities (Curinginnos), license-plate recognition for a national traffic system (GlobalBridge), and a few upstream open-source fixes (Aravis/GStreamer, NVIDIA). The write-ups here are this side of the work.

A note, off the clock

I’ve played electric guitar for over ten years, in just about any genre. Pink Floyd one day, Notorious B.I.G. the next. I can take the lead, or sit back and make someone else in the band the star. Engineering has felt the same to me. Frontend, mobile apps, computer vision, a lot of different fields. The rise of LLMs became the instrument that lets me bring all of those genres together, and point them at someone else’s problem.

Get in touch

Email · LinkedIn · GitHub · Pragmatist · ARKHE