Future Thinking
APPLICATION
Surgical Assistant App
Claude Language Model for Translating predictions into Recommendations.
Hi, I'm Maxence!
My goal is to 1) develop superhuman intelligence that can understand the universe, without optimising on incomplete human knowledge of the world. 2) deploy autonomous decision-making agents given a strict human safety alignment policy.
I'm a PhD candidate at King's College London under the guidance of Prof. Seb Ourselin. Currently, in the SIE-CAI4CAI Lab since 2021. I completed my MSc in Machine Learning and Robitics at University of Surrey and the CVSSP Lab with Prof. Kevin Wells. My work is supported by UKRI's Funding.
View/Download CV | Contact: boelsmaxence@gmail.com
Research
Key Research Areas:
- Surgical World Models: Learning a surgical world model from endoscopic videos to enable general understanding of the anatomical structure and surgical workflows taken by expert surgeons LoViT.
- Surgeon (Agent) Forward Thinking: Now that we understand the agent environment, we need to understant the planned actions and intermediate states towards the desired procedural goal of the surgery SuPRA.
- Temporal Abstraction: While humans plan at an abstract level, control algorithms are still limited by exploring, planning, and assigning credit at the level of primitive actions in progress.
Featured Projects
View All Projects
Short Biography
Maxence is a PhD candidate in artificial intelligence at King's College London, working under the guidance of experts in the field. His research focuses on AI applications for surgical action generation from robotic endoscopic video recordings. His work aims to develop intelligent systems that predict surgical actions by combining general world models with anticipatory decision-making, even in the absence of fully interactive simulations. Maxence is especially interested in integrating 'system 1' and 'system 2' thinking into AI for surgical tasks to enhance real-time decision-making. His previous publications include work on surgical phase recognition using transformers, and he is currently exploring offline reinforcement learning techniques using static datasets. Maxence holds a strong vision for advancing AI in healthcare and is committed to overcoming the challenges of static data environments. He plans to present his latest work at MICCAI 2025, with ambitions of working for world-leading AI labs like Google DeepMind, Meta FAIR, and OpenAI.