AI infrastructure that survives contact with your compliance team.
I design and build the AI products and the infrastructure-as-code behind them — so your organization can move fast on AI without losing track of who’s using it, what it can see, and what happens when it’s wrong.
Most AI pilots die in procurement, not in the demo.
The model works. What kills the project is everything around it — nobody can say who’s allowed to use it, what data it touches, or what happens when it’s wrong. Here’s what that turns into when it’s built right.
A system that has edges, on purpose
Architecture decisions and system boundaries documented arc42-style before the building starts — so what gets built has a shape someone actually chose, not one that happened by accident.
AI features people actually trust
Retrieval-grounded assistants and agentic workflows that answer from your own data — reliable enough to depend on, not a demo that impresses once.
Infrastructure your ops team can own
Everything provisioned as code — OpenTofu, Bicep/ARM, Docker, Kubernetes, CI/CD — so what works in staging is what ships, and your team isn’t dependent on tribal knowledge.
A paper trail, not a black box
Real-time policy enforcement and provider-agnostic routing, so "who can see what" is answerable on day one — not retrofitted after an audit.
Why this works out better than the usual options.
Not a big-agency handoff chain, and not an AI-only specialist who’s never shipped a real product.
How we’d actually work together.
No multi-week sales process. Three steps.
Scope
A short call about your actual problem — not a generic pitch deck. I’ll tell you honestly whether it’s a fit.
Build
I work in your repos, with regular checkpoints — no black-box disappearing for six weeks.
Hand off
Architecture documented arc42-style, owned by your team when we’re done — not something that only works while I’m still around.
Questions people usually have.
Do you work solo or with a team?
What does an engagement usually look like?
Do you sign NDAs?
Can you work with our existing stack?
Do you only do AI work now?
How do we start?
Have an AI project that needs to survive contact with reality?
Tell me what you’re building. I’ll tell you honestly if I’m the right person for it.
Tell me about your project