17 years in software. The last six building products with AI agents as part of the team.
I help founders pick the right shape for their product, hire the first engineers, and still ship code myself with Claude Code every day.
Four parallel Claude Code sessions, SOP-driven agents, decision logs. Architecture gets tested in code the same week, not in slide decks.
Tech lead and architect on a content platform for creators and brands. Two end-to-end subsystems on my desk: regulatory ad-compliance pipeline (~2M creatives a month, zero compliance incidents since go-live) and the multi-tenant team workspace with strict data scopes, so teams inside the same organisation never see each other's tasks, videos or brands. Access enforced at the data boundary, not the UI.
Patient information system for clinics. Doctor finishes an appointment, presses a button. Smart TV in the waiting room shows: who is next, which doctor, route to the room on an interactive map. Plus voice announcement. Running in several clinics in Germany.
Two autonomous Telegram channels. Good News covers tech/AI/startups, Beauty News covers beauty and fashion. 60+ RSS sources, OpenAI filtering, audio digests via NotebookLM, scheduled publishing via GitHub Actions. Fully automatic, fresh content every day with zero manual work.
AI platform for learning and testing. Upload any material, the system generates questions, flashcards, tests. GPT checks answers and explains mistakes. Spaced repetition for long-term memory. Ready to deploy.