Michael Schiffer

I build productive AI systems from the first prompt to production with real users. Currently, seven systems are in operation: RAG with hallucination protection, Agentic Workflows, three-stage OCR pipelines with automated bank reconciliation (>4,000 documents), clinical transcription with direct medical record entry, Decision Support with multi-factor scoring, and a custom workflow platform as an n8n alternative (~24,000 LOC Python/React).

My stack is intentionally self-hosted: Ollama on own GPU infrastructure, FastAPI + PostgreSQL/pgvector as the backbone, Docker-Compose orchestration across multiple servers, Caddy with 7-layer security. Models: Qwen3, Mistral, medgemma, bge-m3 Embeddings. Framework-free because I need control and performance — but with knowledge of alternatives (LangChain, LangGraph).

My background is dual: 20+ years of programming and interim management in automotive, mechanical engineering, and healthcare. Interim COO PWG (350 employees, 80 million EUR), PM at Johnson Controls and Toyota Boshoku, supplier development KTM. Six Sigma Black Belt, DGQ Quality Manager, Certified Program Manager. This means: I translate between business departments, IT, regulatory affairs, and management without buzzword bingo.

Strengths for projects:

• GenAI productionization instead of PoC (data pipelines, LLM integration, API connection)

• Self-hosted & GDPR compliant — relevant for regulated industries (Healthcare, Insurance, Public Sector)

• RAG + Agentic AI with structured hallucination protection through verification

• Process and requirements analysis at a senior level (Six Sigma, Lean, FMEA)

• Bridge between development team, business department, and C-level

Working through The Implementers GmbH.

Native German speaker, fluent in English (US series launch, plant stabilization UK, years of work in US/UK corporate structures).

Available remotely or hybrid. On-site assignments upon agreement.