Nine years building AI where being wrong is expensive.

I have worked on four of the five layers of AI — the compute layer, where the hardware architecture that runs the models is decided; the infrastructure layer, on cloud or edge, which sets the performance-and-power envelope; the model layer; and the application layer, where AI systems are built for a given use case. The fifth is energy, and it is the one I have not worked in directly.

What interests me most now is Physical AI — models that act in the world rather than only produce text. As the models themselves become commodities, the difficulty moves to everything around them: the compute budget on the device, the latency, and the safety argument for a machine that moves. That is the ground I have spent my career on, in cars and in silicon.

At AUMOVIO, formerly Continental, I co-authored the company's first internal AI governance process, aligned to ISO/PAS 8800 and the EU AI Act. What mattered was not the document itself — it was that engineering teams actually followed it on existing customer projects. In the same period I led a patented AI safety architecture through to its commercialisation phase, ran AI and computer-vision SoC strategy for autonomous driving platforms, and worked with a global team on getting models to run economically on hardware platforms such as NVIDIA and Qualcomm. Before AUMOVIO I spent six years at Continental owning the technology roadmap for AI acceleration in cars, and before that a student job at Philips Healthcare training models for MRI and CT diagnostics.

I did my MBA at Frankfurt School of Finance and Management alongside my work, which enabled me to use my technical skills to create and evaluate business value. Most of what stops AI reaching production is not technical — it is that nobody has translated the technical position into something a board, a regulator or a general counsel can act on. Now I work with institutions finding their way through the AI transformation. Usually that means building something specific with them. Sometimes it means showing them what could be achieved, and making them aware of the costs and implications, so they have the clarity to judge whether it makes sense for their business case.