I design AI systems around how people actually decide and act: interfaces that communicate uncertainty, support judgment, and keep humans in control where the stakes are high.
I hold a PhD in Computer Science from Leibniz University Hannover (“Leveraging User-Centered and Neuro-Symbolic Artificial Intelligence in Information Systems”) and have published six peer-reviewed papers on human-AI interaction and knowledge engineering, with more on the way. Behind the research are nine years of applied work at Bosch, eccenca, and MyActivities, building agents, knowledge graphs, retrieval pipelines, and the interfaces that make them usable. I still review and publish, because client work and research hold to the same standard.
Some problems need a simple assistant. Others need several agents working together, remembering what happened, what's known, and how things are done here. I build whichever the problem calls for. Before anyone relies on it, I test it against prepared scenarios: the situations it must handle, and the ones where it has to stop and hand over to a person. Along the way I mentor developers and teach teams what's new, so the know-how stays when I leave.
In 2026, model capability is rarely the constraint; adoption is. AI gets ignored when people can't judge the output, can't recover from errors, or feel slower with the tool than without it. Those are design problems, not model problems.
My process starts with the user's decision: what must this person decide, what evidence do they need, and where should AI propose versus stay silent? Then I design the handoff between model output and human judgment. Get it right and you build AI people rely on instead of AI they bypass.