One person, one invoice.
No consulting team to onboard. I am the architect, the engineer, and the person you call — all in one. The shortest path from question to answer.
Four fields.
One promise:
it runs.
From roadmap to production — I work at every layer of the AI value chain, with a clear price frame and a clear end. No retainer, no rolling subscription — only as long as you actually need me.
§ 01 — What's different
What you get here, and not elsewhere.
Doctorate and production.
PhD in Cognitive Systems plus fifteen years of code. I read the papers that drive the hype — and I build the systems that have to run after the hype is gone.
Fixed price wherever possible.
Discovery, prototype, and workshops at a fixed price. Time-and-materials only where the uncertainty is real — with a weekly budget update, every time.
§ 02 — Core
Four engagement types — in detail.
It runs.
AI Roadmap & Feasibility
Where does AI pay off in your organisation — and where doesn't it? We inventory use cases, assess feasibility and risk, and prioritise by impact. The deliverable: a concrete paper your board can actually read.
- Use-case inventory & scoring
- Technical due diligence (data, infra, skills)
- Build-vs-buy analysis
- EU AI Act classification & compliance check
- 12-month roadmap with milestones
3 – 6 weeks · from €18,000Fixed price
Prototype & Production
From proof-of-concept to production. I build the smallest possible end-to-end slice, harden it, ship it. RAG, agents, fine-tuning — whichever the problem actually calls for.
- RAG pipelines & semantic search
- LLM agents & tool use (ReAct, MCP)
- Fine-tuning (SFT, DPO, LoRA)
- Deployment (on-prem, sovereign cloud)
- Observability, eval & guardrails
2 – 16 weeks · from €1,400/dayFixed price or T&M
Applied Research
You have a research question but no consortium? A consortium but no proposal? I help with proposal writing, partner matching, project coordination, and scientific evaluation.
- BMBF, Horizon Europe, Interreg proposals
- Consortium building & partner matching
- Project coordination (PM, reporting)
- Scientific evaluation & publication
- Technology transfer into deployment
6 – 36 monthsProject-scale
Workshops & Coaching
For teams that need to understand AI, not just use it. Hands-on, on your stack, with your data. No canned slide decks and no generic exercises.
- Executive briefing (½ day, remote)
- Engineering bootcamp (2 – 5 days, on-site)
- 1-on-1 coaching for tech leads (monthly)
- Interactive code reviews with your team
- Supervised pilot projects as a learning vehicle
½ – 5 days · from €2,800Fixed price per module
§ 03 — Process
How a project runs with me.
Phase 01
Listen.
Two half-days on-site. I ask, you talk. The outcome: a list of possible entry points, sorted by effort and impact.
Phase 02
Prototype.
The smallest possible end-to-end slice. Real data, real code, real UI. No wireframes, no click-dummies.
Phase 03
Harden.
Evaluation, edge cases, security, monitoring. The prototype becomes a system. Demo becomes production.
Phase 04
Hand off.
Documentation, handover workshops, 3 months of standby. You own it now — with me as backup.
Evaluation · Guardrails · Handover · Ship
§ 04 — Specialisations
Where I'm particularly called in.
Evaluation that convinces.
How do you know your LLM system actually delivers what it promises? I build eval frameworks on real user scenarios, not just benchmark numbers.
Sovereign & on-prem.
When the data can't leave your servers: Llama, Mistral, Qwen on your own infrastructure. Including eval pipeline and operations plan.
Research → product.
You have a research prototype and don't know how it becomes a product? That exact crossing is my craft.
§ 05 — Straight answers
Three questions I always hear.
q_01.md
"Can't you just send a proposal?"
No. I need 30 minutes with you before I can tell you whether and how I can help. Anything else would be guessing. And guessing is exactly what breaks AI projects.
q_02.md
"Do you work with enterprises?"
Yes — but only with a clear internal sponsor and a procurement team that isn't building a framework monster. I turn down 90% of enterprise enquiries. The rest, I ship.
q_03.md
"Guarantee it works?"
Guarantee a specific outcome from a probabilistic system? No. But: clear eval criteria, weekly transparency, and a contractual exit after every phase if it doesn't pan out.
§ 06 — Let's start
Does one of these fit your project?
In the first call we figure out which of the four directions matches your topic — or whether it's a blend. No obligation, 30 minutes, German or English.