2026 DIA Global Annual Meeting: Why Federated Learning Reduces Friction

Full interview summary

In a recent video interview with Applied Clinical Trials at the 2026 DIA Global Annual Meeting, Ittai Dayan, co-founder and CEO of Rhino Federated Computing, discussed why data fragmentation remains one of the most persistent barriers to meaningful AI application in clinical trials and real-world data use, and how federated learning addresses—and does not address—that challenge. He opened by describing the scale of the fragmentation problem: data sits siloed across companies, geographies, modalities, and regulatory regimes, and assembling enough of it to support serious AI development requires coordination across a large number of collaborators whose participation is shaped by trust, incentive alignment, and perceived risk. He noted that the United States is an outlier globally in having invested substantially in both data sharing frameworks and infrastructure, while clinical trial and discovery data in most other regions remains significantly more fragmented.

Dayan was careful to draw a clear distinction between what federated learning does and does not do. It reduces friction—eliminating the need to centralize data and removing many of the consent, IRB, and data pooling concerns that slow collaboration—but it does not manufacture the incentive to share in the first place. That incentive must already exist. Where federated approaches are most powerful, he argued, is in cases where combining data creates new value that feeds back to contributors, citing Eli Lilly's TuneLab as an example of a model where biotechs share data and receive improved AI applications in return.

He closed by outlining what sponsors actually need for federated approaches to work at speed: program-level deployment rather than organization-wide rollouts, enterprise-grade legal and security infrastructure, and technology that meets researchers and data scientists where they are—without requiring software engineering expertise to operate.