Press & founder story.

I build systems where being wrong isn’t an option. For fifteen years I’ve worked in fault-intolerant, regulated domains — industrial-control infrastructure at Schneider Electric, healthcare AI at Overjet, hospital cybersecurity at Tausight, FedRAMP-High endpoint security at SentinelOne. The thread through all of it: when the stakes are real, “probably correct” isn’t good enough. You don’t sample and hope — you test the whole system, prove every claim, and get the same answer every time you run it. Reproducibility isn’t a feature; it’s the job.

When I looked closely at how audit and assurance actually work, I saw the opposite. Auditors test samples, not populations, and live with the sampling risk that leaves behind. The work rests on judgment that can’t be reproduced — run it twice, get two stories. And a wave of AI tools is making it worse, not better: they feed sensitive client evidence into black-box models that can’t give you the same answer twice and can’t show their work. For an engineer trained to prove things, that was alarming. We were bolting non-determinism onto a profession whose entire product is trust.

So I asked a different question: what if you brought that discipline to assurance? Test the full population, not a sample. Make every result bit-identical when you re-run it. Never let client evidence touch the model — use AI only to author the test from a de-identified description of the data, then execute that test deterministically. Keep the auditor in control, with a clean chain of custody. That’s Luca.

The name is deliberate. Luca Pacioli gave the profession double-entry bookkeeping more than 500 years ago — rigor as the foundation of trust. Luca is that same idea for the AI age.

My path here wasn’t linear — I taught myself to code in Beijing, came back to formal computer science at CU Boulder, and have spent my career working adjacent to a domain’s core function: the architecture, not the trade; the standard, not the operation. That structural vantage point is what lets me encode a regulatory canon into software faithfully.

And I’m building it in Durango, Colorado, on purpose. Rural America doesn’t have to wait for technology to be handed down from the coasts — it can build it. A venture-scale AI company in the mountains is exactly the kind of thing this place can produce.

The north star is simple to say and hard to earn: an audit gives reasonable assurance, never absolute — and the profession is right about that. But for the controls we can test deterministically across the whole population, we’re closing that gap. Not with a slogan. With a method. We’re early, building alongside design partners — but the engine already does what it was built to do.