Speaker
Description
HPC centers are evolving rapidly. In addition to traditional simulation workloads, they are now expected to support AI training and inference, interactive research environments, and increasingly data-intensive scientific workflows.
This keynote looks at how the AI Factory model is reshaping the way these infrastructures are designed and operated. Slurm, Kubernetes, virtual machines, bare metal, high-performance storage, accelerators, and AI services increasingly need to coexist and share the same underlying resources.
Drawing on experience from European AI Factory initiatives, the session will discuss the practical challenges of operating this new generation of infrastructure: how to support very different workload models, allocate resources efficiently, preserve performance, and avoid creating isolated technology silos.
The keynote will also explore the growing convergence of HPC, Cloud and AI, and the role that open platforms can play in building flexible, interoperable scientific infrastructures in Europe.