Speakers

Marc Boubnovski Martell is a Senior AI/ML Scientist at Novo Nordisk and Honorary Research Fellow at Imperial College London, where his research sits at the intersection of structured machine learning and life sciences discovery. His work focuses on designing learning systems grounded in physical and logical structure — ensuring that learned representations remain reliable under noise, anisotropy and domain shift, rather than incidental to the training data. 

At Novo Nordisk, Marc develops large language models, biomedical knowledge graphs and intelligent agents to accelerate drug discovery and translate biodata into scientific insight. His contributions include establishing foundational models for structuring unstructured biomedical text via an in-house engine, architecting an industry-leading knowledge graph integrating literature with public biomedical resources, and pioneering agentic AI frameworks capable of complex multi-step biological reasoning. As an author and co-author of a series of publications on perturbation prediction, he also contributes actively to the broader scientific community's understanding of how AI can model cellular responses to biological and chemical interventions. He also builds an evaluation and benchmarking platform that enables transparent model assessment and fosters scientist trust in AI-driven workflows. 

His current research emphasis is on agentic LLM systems that combine structured search and reinforcement learning including Group Relative Policy Optimization (GRPO) to navigate complex biological manifolds and reason over literature, patents and biomedical data. 

Trained in Mathematics and Physics at the University of Glasgow's School of Physics and Astronomy, Marc's earlier academic work at Imperial College London spanned 3D medical imaging, multimodal representation learning and generative models, with applications including CT super-resolution, virtual biopsies for disease analysis, and deep learning approaches linking chemical with imaging data for prognosis prediction. 

A regular contributor and presenter at premier venues including NeurIPS and ICLR, Marc actively shapes the frontier of AI research across both industry and academia