Portfolio

AI applications and the data behind them

Start with my RAG learning assistant to explore retrieval, guardrails and LLM evaluation. The data projects below show the Python/SQL pipelines, validation work and spatial analysis that I bring to AI engineering.

AI engineering

RAG learning assistant

I built a learning assistant to compare baseline RAG with layered guardrails. The question behind it is practical: which checks help the system answer from course material, and which ones get in the way of a valid request?

Try the deterministic offline demo, or inspect the BGE-M3/Chroma retrieval path and OpenAI-compatible model stages in the code. The local review tools let you examine evaluation data, with calibration kept separate from final evaluation.

Data engineering and analysis

These projects cover public-data ingestion, reporting datasets and checks on spatial joins. NRW Charging Atlas brings that work into an application for charging operators investigating where to expand. You can also read how I built it.

Pinned projects

Related material

Read a longer project story, open a slide deck or follow the references.