At Ray Summit 2025, Dmytro Spodarets sat down with Keerti Melkote, CEO of Anyscale, to discuss the newly announced Anyscale Azure integration and the role of Ray in modern AI infrastructure. Keerti explains how Ray is used to power training, inference, multimodal data processing, and emerging agentic workloads, giving enterprises a flexible foundation for building AI systems at scale.

Keerti also shares why Ray’s transition into the PyTorch Foundation strengthens its position as an open and community-driven technology, and how the Azure integration improves security and operational consistency for enterprise teams. He highlights practical challenges companies face today — from GPU shortages to the need for dynamic resource allocation and reliability when running distributed AI workloads — and discusses how Anyscale is addressing these issues.

This interview offers a clear and practical look at the future of AI infrastructure, with insights into how organizations can evolve their architectures, overcome scaling bottlenecks, and prepare for the shift toward AI-native systems.