Senior Python & AI Engineer in Hamburg. Nearly 10 years taking software from idea to production — backend, data platforms, cloud — building with LLMs since 2022.
By day I build the data platforms behind energy systems and migrate legacy pipelines with agentic coding and human-in-the-loop review. Outside work I build agentic RAG systems and measure them with evaluation sets, then write about what I learn.
| Project | What it is | Stack |
|---|---|---|
| InsightOS | Hybrid RAG over a Notion knowledge base: pgvector + pg_trgm with Reciprocal Rank Fusion, SQL metadata filtering, confidence-based routing across three open-weight model tiers, source citations, cost/latency telemetry, eval set as release gate | LangGraph · FastAPI · PostgreSQL 16 · Hugging Face · Chainlit · Docker |
| Deutsch Tutor RAG | Grounded agent that answers only through an explicit retrieval tool, returns validated structured output, says "not found" instead of hallucinating; PII redaction, model fallback, human approval before KB changes | LangChain · Qdrant · Pydantic · OpenAI · Docker |
| generative_ai_portfolio | Generative AI experiments and learning projects | Python |
Agentic AI & RAG: LangGraph, LangChain, Claude Agent SDK, agent skills, tool calling, structured output, hybrid retrieval, pgvector, Qdrant, evals (Hit@5), LangSmith Engineering: Python (FastAPI, Pydantic, SQLAlchemy, Pytest), PostgreSQL, Rust, Go Data & platform: Airflow, data lineage, data quality, AWS, Azure, GCP, Docker, Kubernetes, Terraform, CI/CD Domain: EU power markets and energy systems data
LinkedIn · Hamburg, Germany · English (C2), German (B1 in progress)