Accelerating Discovery While Protecting Patient Data: Federated Learning at Scale

Federated learning is a machine learning approach that trains models across decentralized datasets without pooling patient-level data into a central repository. For biotechnology executives, it can accelerate multi-site discovery, improve model generalizability, and reduce data-transfer friction. For patient groups, it offers a way to enable research collaboration while strengthening guardrails around trust, consent expectations, and stewardship. This panel brings together a patient-organization leader, industry R&D data executives, and multi-institution collaborators to break down (1) how federated learning works in practice, (2) when it is the right tool versus alternatives (e.g., pooled data, synthetic data, secure enclaves), and (3) the governance and operating model required to scale beyond pilots. Speakers will share real-world case examples (e.g., neurodegenerative and other diseases) including partner roles, data readiness, model validation, bias and equity considerations, and what “success” looked like for both patients and product development.