I’ve worked on AI systems across a range of domains, from in-car infotainment and educational technology to scientific publishing and biomedical applications. Projects have ranged from semantic search and retrieval-augmented generation to information extraction and agentic research workflows. I enjoy researching and figuring out how to make AI useful in uncertain, novel application areas and difficult use cases. After all, who can resist a good technical riddle?

In my academic life, I researched AI systems in education at TU Darmstadt. My dissertation investigated how to improve reading comprehension with automatic question generation during the advent of modern transformer-based models. Being in academia when modern AI took off resulted in all kinds of fun research: from fine-tuning models and architectures to figuring out boundaries and evaluation scenarios and adversarial attacks in educational domains. I am a firm believer that AI could greatly transform education by improving a lot of the non-fun parts of teaching. I am also a firm believer that the inertia of the German educational system will make this so much harder than it could be.

Here I write mostly about AI: what works, what doesn’t, and what I haven’t quite figured out yet. Away from a screen, I’m usually climbing, hiking, or reading, ideally not all at once.