For about twenty years I worked in academic project management and administration — graduate schools, research institutes, controlling. Along the way I kept running into the same small, unglamorous problems: booking a shared resource, tracking a deadline, calculating personnel costs, pulling structured data out of a PDF someone sent as a scan.
I started building small tools to solve these myself, usually with AI-assisted coding, and always starting from a real, specific need rather than a tutorial.
Some of the things I've built:
- 🗓️ gutachten-dashboard — deadline tracking and process KPIs for expert-report workflows, with access control and full change history
- 💶 gehalts-guide — a calculator for German public-sector personnel costs (TV-L / TV-Ärzte)
- 📄 pdf-key-extract — client-side extraction of structured data from PDFs, including scanned pages
- 📝 format-fantastic — import any CV as PDF, auto-detect its structure, edit freely, export print-ready
- 🔬 research-output-monitor — a Python tool that tracks new PubMed publications against a group's website and flags what's missing
- 🚗 shared-drive-pal — booking and handover tracking for a team-shared car
- 🖥️ buddy-desk-finder — a quiet little dashboard for booking desks and seeing where teammates are sitting
- 💬 ask-my-portfolio-ai — a small RAG chatbot that answers questions about a personal essay collection, with sources
I also write essays and reportage (InZeitung, Schopf2 Poetry Magazin) — a different kind of structure, but structure all the same.
Currently curious about where careful process work meets practical AI tooling — including a small side experiment comparing TabPFN to classical baselines on a public dataset, mostly to understand how it thinks.