I am a full professor of Human-Centred Transformative AI at Hasso Plattner Institute in Potsdam and a Research Consultant at Google DeepMind. I work on the foundations and safety of frontier AI. My team and I focus on understanding and controlling what advanced models learn from data. Our key research areas are:

• AI reasoning, post-training and monitorability
• Data attribution
• Continual learning, test-time adaptation and unlearning
• Training dynamics, memorisation and learning theory
• Multimodal models

Before working full-time on AI, I trained as a physician and radiologist. This background continues to inform some of my work on reliable AI in high-stakes environments.