01 — Start here
"I'm getting started and want to see real work, fast."
If we haven't worked together yet, these are the two ways in. One takes a single day. The other is a full build. Both end with working code your team can run.
1 day
One day. I take a real problem your team is facing and build a working prototype of it in front of you, talking through each decision as I go. No commitment beyond the day. Your team leaves with the code and a clear sense of how I work.
3–5 weeks
Your first real AI system in production: knowledge retrieval, document processing, a structured assistant — whatever the actual need is. I work directly with whoever understands the problem best: the researcher, the operations lead, the founder. You describe what needs to happen in your terms. I translate that into engineering. No coordinator needed between us. Built alongside the people who will own the system, so the code and the confidence to run it both stay when I leave.
02 — Already building
"I have something that works in a demo but stalls before production."
When an AI system gets stuck on the way to production, the cause is usually a mix of architecture, prompts, and a few things nobody secured. This is a focused look at what's holding it back. It only makes sense once you already have something built, so it's rarely a first step.
1 week
I go through the system you already have: architecture, prompts, retrieval, and how your AI agents are deployed. What works, what will break under load, where data could leak, and what to fix first. You get a clear, ordered list your engineers can act on. Part of this comes from doing the same review for a research institute running agents on confidential data.
03 — Ongoing
"Something's live, and I want a senior perspective on call."
Once a system is running, the questions keep coming: a vendor to weigh, an architecture call to make, a hire to assess. This is for clients I've already delivered for. It's what the work grows into once the first project lands.
or €200/hr
A senior technical perspective on call for your AI decisions: architecture reviews, vendor choices, the questions your team can't answer yet. Five to eight hours a month on retainer with priority access, or by the hour when you're not ready to commit monthly.
Who this is for. This works best for organisations where someone inside already understands the problem well. You don't need a technical team or a coordinator between us — I work directly with whoever holds the knowledge, whether that's a researcher, a department head, a founder, or an operations lead. You explain what needs to happen; I translate that into something that runs. Usually a startup, a research group, a mid-sized company, or one department inside a larger one. If you're after a company-wide AI strategy with a big-firm name behind it, I'm probably not the right fit, and I'll tell you so early.
How payment works. Engagements are project-based, not hourly. Build work is split across milestones: 30% to start, 40% at the mid-point, 30% on completion.
The review guarantee. If a Production AI Review gives you nothing you can act on, you get your money back. The success criterion is written into the contract before any work starts.