Announcement_ddsa_fellowship

I am honoured to have been awarded a DDSA Postdoctoral Fellowship for my project Geometry of Trust: A Unified Framework for Convex Latent Steering and Causal Concept Alignment, starting September 2026, based at the Section for Cognitive Systems at DTU Compute. The project asks whether AI safety can come from inside the model — from the geometric structure of how it represents information — rather than only from filtering inputs and outputs, building on our Nature Communications paper.