Fifteen years of
production engineering.
Most AI consultants come from ML. I came from production systems. That's the difference.
I spent fifteen years building production systems at companies like Vimeo and Zumba — shipping features that moved business metrics, leading teams through platform migrations, and solving hard problems under real constraints. I was good at it. But I wanted more direct ownership of the work, autonomy in how it got built, and a closer line of service to the people I'm building for.
So I went out on my own.
The motivation
The most consequential AI work is happening inside enterprises, where a system has to earn trust from engineering, security, and the people who rely on it. Bringing production engineering experience directly to those teams felt like the right way to be useful.
Production AI is a systems problem
Most AI failures aren't model failures. They're systems failures — bad context management, missing guardrails, no monitoring, no plan for how real people will trust, use, and abuse the thing.
Shipping past those failures takes the kind of systems thinking you get from fifteen years of production work: pattern libraries, failure-mode intuition, and the judgment that comes from putting real systems in front of real users. Years in the trenches is a feature, not a liability.
The work
I embed with enterprise teams to understand their domain, constraints, and workflows — then build AI and automation systems that move real business KPIs. Increased revenue. Reduced costs. Faster operations. Doing this well requires that embedded element — you can't automate what you don't deeply understand.
I own and maintain everything I build as a long-term growth partner.
Who I work with
Enterprise engineering teams with AI prototypes that have not made it to production, or AI on the roadmap without a clear path to build systems their customers, operators, and security teams can trust.
Common signals:
- A working prototype that's been stuck in “almost ready” for 60+ days
- AI features on the roadmap, no in-house engineers experienced in shipping LLM-powered systems to production
- Engineering hours eaten by repetitive workflows that should be automated
- A demo that's impressive in front of users but won't survive real load, real costs, or real edge cases
I also take selective work outside this profile when the problem fits. But the work I'm built for is helping enterprise teams earn trust in AI systems before those systems become business-critical.