AI that holds up in production
AI reaches production and underdelivers for concrete reasons. Requirements get missed across both the technology and the implementation process. The model ships without the tooling that would make it useful in real work. And it gets evaluated like deterministic software, when its behavior is probabilistic. I work across that, from strategy to secure architecture to the build itself.
How I work
My engagements start from a conversation, often about something I've published that maps to a problem you're already facing. We talk about how it applies to your systems, and the work gets scoped from there.
I work alongside your engineers, IT, and security. I can lead the build or embed with your team and hand it off documented and clean.
Read the thinking→The work, in public
Recent writing on the problems I get hired to solve. The same thinking I bring to an engagement.
Ways to work together
Defined engagements for smaller, scoped work. Larger builds start from a conversation.
Production Build
A real system in users' hands.
I lead the end-to-end build of an AI system that holds up in production, from secure architecture through implementation to a clean handoff.
Start a conversation→Service 2Business Automation
Stop spending hours on tasks that should take minutes.
Custom automation workflows using AI and your existing tools. Reporting, document processing, data workflows, client onboarding. If it's repetitive, it can be automated.
Start a conversation→A conversation about a specific problem
From there I scope the engagement to what's in front of us, and you'll see clear deliverables and price before anything begins.
Start a conversation