From idea to production, no detours.

No, and I would rather say it here than on the third exchange. I am starting my practice under my own name. The systems described in my work were built on engagements, under contract, and I can name neither the clients nor their internal numbers. What I can do instead: open the code and the architecture, walk you through the trade-offs I made and why, and let you judge on what I write publicly. And start with a four-day audit, which commits you to nothing beyond it.

The durations are public: audit 4 days, POC 12 days, build from 20 days, embedded architect 10 days minimum. The firm price is set in the quote, once the scope is clear, and you get a ballpark from the first exchange, for free. I don't publish a rate card because it would be false: two projects of the same type don't carry the same budget depending on the state of your data and your reliability bar.

On a fixed price, the overrun is on me: you know the price up front and I carry the risk. If the scope changes along the way, that is an amendment signed before the spend, never a surprise invoice at the end. On time and materials, the committed volume and the end date are written into the contract. Either way, I tell you before, not after.

Yes. The code, the architecture, the documentation and the audit report are yours. Standard architecture, open standards like MCP, handover included in every build: any competent developer can pick it up after me. Once it ships, your team keeps control, which is the point. If you still need me, it is on time spent, under an agreement signed at delivery. I don't sell subscriptions.

Yes, and you pick the level. Depending on your constraints: EU hosting, open-source models deployed directly inside your own infrastructure, or cloud under a strict agreement. On-prem, your data never leaves your perimeter. In every case your data is never used to train public models, and I will sign an NDA from the first conversation if you want one. We weigh the cost, performance and sovereignty trade-off together.

Poorly architected, yes. The right model in the right place, caching and batching change everything: the order of magnitude targeted is 30 to 70% savings at equal quality, costed against your own bill during the audit. And I only build the cases that pay off.

No. A lot of repetitive work is solved with a simple, well-designed workflow: cheaper and more predictable. I only add AI where judgment is needed: classifying, extracting, drafting. It's often the best place to start.

That's the real issue. Sourced answers (RAG), guardrails on outputs and evals on every version: I measure quality before production and stop drift instead of just living with it.

An audit is measured in days, a POC on your data in a few weeks, a production build in weeks to a few months depending on scope. From the first conversation, you have a clear timeline.

I'll tell you honestly. An audit quickly separates the high-ROI cases from the ones where AI adds nothing. An honest no beats a project that disappoints.

Yes. I take over what exists, make it reliable (evals, guardrails, observability) and move it from demo to production, or bring its bill down.

It's planned for from day one: standard architecture, documented code, handover included in every build. The system belongs to you, and any competent developer can take it over. Working solo also comes with one guarantee: the person who designed your system is the one maintaining it.