codon.Applied ML since 2012codon.Applied ML since 2012You’ll pay for a learning curve either way. Mine’s already paid. Eighteen systems in production across twelve organizations, cancer genomics then, agentic systems now. I own your technology function, build the team, and hand you the keys.
Eighteen systems in production across twelve organizations, with value shipped inside one to two quarters, in a market where most AI pilots never reach production at all.*
* MIT NANDA, The GenAI Divide (2025): 95% of enterprise pilots show no P&L impact · S&P Global Market Intelligence (2025): organizations scrapped 46% of AI proofs-of-concept before production
Fractional CTO is the lead engagement; the others are narrower doors into the same model. Each ends the same way. You own a team that ships faster than the companies you compete with, and it keeps shipping after I’m gone.
If it isn’t decreasing, something is wrong. Conversion fees decline to zero. The longer you keep an engineer, the cheaper they are to hire. I’m the only vendor whose plan is to become unnecessary.
Both times it reached production. In my experience the blocker is rarely the technology. It’s that nobody owned whether the output could be trusted once real decisions depended on it.
I’ve been applying machine learning to hard problems since 2012. Deep learning on cancer genomics at Washington University first, then AI functions built from zero four times: Domo, Allurion, Falcon, and a national medical records service.
I’ve been applying machine learning to hard problems since 2012, starting with deep learning on cancer genomics at Washington University School of Medicine. That work is published and dated: first author in Nature Genetics and Nature Methods, co-author in Nature.
That isn’t a credential for its own sake. It’s a record of building models people stake real decisions on: cancer genomics, clinical records, risk prediction. In that world, evaluation, guardrails, and human-in-the-loop design aren’t afterthoughts. They’re the job. Everyone can build a demo now; that stopped being the hard part. Knowing whether you can rely on the output is the whole thing.
I’ll say the obvious thing before you have to ask: I haven’t held the CTO title at a company that employed me full-time. I hold it fractionally for multiple clients right now, owning the function and the team at each. At Domo I ran a business unit end to end: architecture, delivery, hiring from top programs, product, product marketing, and the sales engineering that sold it. At Allurion I owned the backend and data platform. And Codon has been profitable in 2024, 2025, and 2026, so hiring, margins, and contracts aren’t theory. If you need someone who has scaled a two-hundred-person engineering organization, that’s a different person and I’ll tell you so. If you’re hiring your first real technical leader, this is the experience that matters.
Tell me where you are. I’ll tell you straight whether I can help, and how. If it isn’t a fit you’ll hear that too, quickly.
Send this and you’ll go straight to my assistant’s calendar to pick a time.
hello@codon.llc
