AI Engineer · Switzerland

AI Engineer from Switzerland — Claude, OpenAI, MCP, RAG, Agents

I build AI into products people actually use. Claude (Anthropic) specialist with Certified Architect credentials, MCP tooling, OpenAI integrations, RAG pipelines, and autonomous agents. 10+ years of production engineering discipline applied to LLM systems — so your AI feature does not become a liability.

Available for AI engagements worldwide

Claude Certified Architect

Deep expertise in the Claude Agent SDK, Claude Code, the Anthropic API, prompt caching, MCP, and agent workflows.

Production discipline

Observability, evals, rollback strategies, cost controls, zero-retention options. AI that ships without 3am incidents.

Swiss data residency

EU-hosted deployments, PII redaction, audit trails, and contractual clarity for regulated environments.

01.What I deliver

Claude & OpenAI Integrations

Drop LLM-powered features into your product: summarisation, classification, structured extraction, semantic search, chat assistants.

Custom MCP Servers

Expose your internal tools, databases, or APIs to Claude Code, Claude Desktop, and other MCP clients. TypeScript or Python.

RAG Pipelines

Production-grade retrieval with embeddings, reranking, chunking strategies, hybrid search, and evaluation harnesses.

Autonomous Agents

Multi-step AI workflows using tool-use, the Claude Agent SDK, LangChain or custom orchestration. With guardrails and observability.

LLM Evaluation & Hardening

Prompt-injection defence, eval suites, A/B of model versions, cost and latency budgets, incident playbooks.

AI Product Strategy

Pre-build workshops: what is actually a good LLM use case, what will embarrass you, what will pay for itself.

02.Process

  1. 01

    AI audit (free, 60 min)

    We look at your use case, your data, your users. You leave with a clear build vs. skip recommendation.

  2. 02

    POC or scope quote in 1 week

    Either a fixed quote for a Build engagement, or a 1-2 week AI Sprint to de-risk assumptions with a real POC on real data.

  3. 03

    Ship with guardrails

    Evals, cost alerts, fallback paths, content filtering, input validation. AI features that are boring in the best way.

  4. 04

    Monitor & iterate

    Model upgrades (e.g. Sonnet 4.6 → 4.7), prompt tuning, RAG quality improvements, usage analysis. Available as retainer.

03.Packages & Pricing

AI Sprint

from CHF 3,900

1 to 2 weeks

POC, integration, or MCP server. De-risk your AI idea on real data before committing.

AI Build

from CHF 8,900

3 to 6 weeks

Production LLM feature or full AI product. Evals, monitoring, cost controls included.

AI Retainer

from CHF 1,800 / month

Monthly

Ongoing AI engineering capacity. Model upgrades, prompt tuning, incidents, quality work.

04.FAQ

Claude or OpenAI — which do you recommend?+

Depends on use case. Claude for long-context, complex reasoning, tool use, and when you need Swiss/EU data residency via Anthropic's EU deployment. OpenAI for broadest ecosystem and some speech/vision tasks. I regularly build with both.

Can you train a custom model?+

For most business use cases, fine-tuning is the wrong answer. RAG + good prompting + small focused evals wins 90% of the time at a fraction of the cost. If fine-tuning is genuinely needed, I handle that too.

Do you deploy on-premise / air-gapped?+

Yes, via Anthropic ZDR, private OpenAI deployments, or open-weight models (Llama, Qwen, Mistral) on your infrastructure.

What about prompt injection?+

Every production LLM deployment I ship has input sanitisation, output filtering, tool-use allow-lists, and a documented threat model. Not an afterthought.

Can you integrate with Claude Code or MCP?+

Yes. Including custom MCP servers, Claude Code skills, and agent SDK workflows. Claude Code is part of my daily workflow.

What is the minimum engagement size?+

AI Sprint at CHF 3,900. Below that, the overhead of doing it properly does not amortise.

Ready to start?

Tell me what you are working on. I usually reply within 24 hours.

Start AI Project