AI Tool ConsultantNotes from real AI builds
Long-form write-ups of the projects behind the skills: what worked, what broke, and the rules that came out of it.
Best AI Stack for Building on a Budget
What I learned building myberuf with Claude, Codex, ChatGPT, Mistral, Vercel, GitHub, project docs, smoke tests, and a stubborn respect for small vertical slices.
Read → field note / usable privacy / not legal adviceUsable Privacy in AI
Privacy in AI cannot be reduced to longer policy text. It has to be designed into data flows, model choices, interaction patterns, and operational habits that still let users complete their real task.
Read → case study / live MVPmyberuf.com
A German-for-work learning product built around workplace situations, AI coaching, mistake review, progress tracking, and handover-heavy AI-agent development.
Read → case study / local LLM evaluationGOJA: local LLMs for structured extraction
My Master's study project (summer term 2026) asked whether a general-purpose language model, running locally, can reproduce a fine-grained annotation scheme for German job ads from instructions and two examples alone. The answer: useful for model evaluation and human-reviewed pre-annotation, not a replacement for the supervised model.
Read →German readers: see also Datenschutz and AI.