Jeremy's dissertation was co-supervised by Benjamin Kuipers and Risto Miikkulainen. It focused on problems in bootstrap learning and developmental robotics, specifically methods that agents can use to learn about the geometry of things with little or no prior knowledge.