Maxence Boels

Research Interests

I've spent my PhD focused on a core problem: ``how to make surgical robots understand long-contexts like multiple hours videos to foresee long-horizon tasks and predict abstract goals into concrete actuations.''

Key Research Areas:

• Surgical Video Understanding: Recognising surgical anatomical structures and procedural steps—instruments, organs, gestures, and phases— from long and complex surgical videos using deep learning models [ LoViT ],
• Expert Imitation Learning: Generating and mimicking expert-level surgical workflows by learning from expert demonstrations using video recordings for planning and decision-making in surgical procedures [ SuPRA | SWAG ],
• Workflow Discovery and Planning: Selecting optimal next surgical actions through reasoning and planning about expected future rewards using reinforcement learning, and discovering optimal surgical workflows that maximise desired procedural outcomes for patients, searching for new pathways beyond expert demonstrations [ DARIL ].

Short Biography

Maxence is a postdoctoral research associate in the Prorok Lab at the University of Cambridge's Department of Computer Science and Technology, working with Amanda Prorok. His research focuses on vision-based navigation and collaborative intelligence in multi-agent systems. Previously, Maxence completed his PhD at King's College London, specializing in surgical video understanding, expert imitation learning, and workflow discovery and planning. He has a strong background in machine learning, computer vision, and robotics, and has published in top-tier conferences and journals including MICCAI and IJCARS.