Work with me
Most AI projects stall between a demo and something a team can rely on. I take it from one to the other, and I stay for the part after launch.
What I build
- AI workflows that run a real process. An agent or pipeline that does a job end to end, for example triaging requests, drafting documents or reconciling data, with a person approving the steps that matter.
- MCP servers and tool access. Your internal systems exposed to AI agents with authentication, per-tool permissions and an audit log.
- Serving and operations. Model hosting on Kubernetes, tracing, cost per request, rollback. The same standard as any production system.
- Evaluation. A test suite for AI behaviour that runs on every change, so quality cannot drift without anyone noticing.
Making your team the supervisors
AI does the first draft; people decide. I run short working sessions that teach a team to review AI output, write good instructions, spot failure modes and know when to stop a run. This also covers the AI literacy duty in Article 4 of the EU AI Act. After launch, an optional run plan keeps a named person watching quality, cost and incidents.
How we start
A free 30-minute video call, then a short written plan: what to automate first, what it needs, what it will cost. Small first step, checked by you before we go further.