PortfolioBlogAboutContact

AI Integration Consultant for Swiss Teams

AI integration consultant for Swiss teams

Anthropic SDK, OpenAI SDK, production AI features. EU-incorporated, CHF or EUR invoicing. Async-first.

Why most AI projects fail in production

The demo works. The chatbot answers correctly. Then real users, real edge cases, real cost budgets, real compliance obligations. "It worked locally" is where most AI projects die. In Switzerland, add data residency, FADP/DSG compliance and (often) German-language quality concerns to the failure modes.

I build AI features that survive production. Anthropic and OpenAI SDKs, prompt caching, structured outputs, tool use, evals, cost tracking. Real client projects, not demos.

What I do

  • AI feature integration into existing applications, with proper evals and observability.
  • Prompt engineering with measurable tests. Not vibes.
  • Workflow automation for internal tools that save your team hours per week.
  • RAG and semantic search with vector databases.
  • Cost optimization. Prompt caching, model selection, batching.
  • Compliance-aware deployment. Data residency, FADP/DSG considerations, EU-hosted inference where required.

Proof, not promises

  • Flowrence (healthcare): production architecture handling sensitive regulated data, relevant for AI features in compliance-bound contexts.
  • Gems XYZ (Web3 social): real-time systems with external API integrations.

I also run the Skool community "The Agentic Architect Lab" for technical founders shipping AI features.

How engagements work

  • Discovery first: paid 1-week scoping engagement before any build.
  • Then build sprints, two to eight weeks.
  • Invoice in CHF or EUR. VAT reverse-charge for Swiss VAT-registered clients.
  • Async-first with weekly sync.

FAQ

Anthropic, OpenAI, or open-source? Often a mix. Claude for reasoning, GPT for structured tasks, open-source for high-volume cheap inference. Recommendation based on your case, not vendor loyalty.

Data residency: can we keep AI inference in the EU? Yes, with the right model choices. Anthropic and OpenAI both offer EU endpoints; self-hosted models are also an option if FADP/DSG concerns warrant it.

How do we know the AI is actually working? Evals. Test cases run on every prompt change. Without evals, you're guessing.

Contact

If you'd like to discuss a project or just figure out whether we're a fit, get in touch via the contact page. No sales pitch.