I am excited to share that I will be speaking at the upcoming NeurIPS 2026 Workshop on World Models for High-Stakes Health (WMHS) in Atlanta, Georgia on December 12, 2026.

Healthcare has leveraged machine learning to model disease progression, treatment outcomes, and operations for decades. However, the emerging frontier is whether world models can reliably generalize across heterogeneous patient populations, complex interventions, clinical workflows, and diverse care settings—while remaining rigorous and precise about uncertainty and the deep biological unknowns inherent in clinical data.

“For me, the key opportunity is not simply better prediction. It is creating systems that continuously learn from decisions, observations, outcomes, and feedback. If successful, world models could become a foundation for patient simulation, clinical development intelligence, and more adaptive healthcare systems.”

The workshop convenes leading researchers and practitioners across machine learning, clinical development, and trustworthy AI to explore:

• Patient World Models: Learning from longitudinal, multimodal clinical data and patient timelines.
• Clinical Trial Simulation: Virtual trial arms, synthetic controls, and external control cohorts to de-risk development.
• Intervention-Aware Reasoning: Counterfactual outcome prediction, causal representation learning, and treatment effect modelling.
• Continuous Learning Loops: Closing the gap between predictions, decisions, real-world observations, and model refinement.

I look forward to discussing both the immense opportunities and the open challenges with an exceptional community of collaborators and organizers from Novartis, IQVIA, GSK, AstraZeneca, Microsoft, NVIDIA, University of Toronto / Vector Institute, Columbia, and Oxford.

For more details on the program and call for papers, visit the WMHS @ NeurIPS 2026 website, or join the discussion on LinkedIn.