Explore the capabilities of Kueue, a cloud-native job scheduler for building multi-tenant batch systems on Kubernetes clusters, in this informative conference talk. Learn about Kueue's architecture, extensibility, and its ability to support various workloads while implementing job queueing based on quotas, priority, and resource sharing hierarchies. Discover how Kueue operates in both on-premises and autoscaled cloud environments, maximizing resource utilization through borrowing and preemption mechanisms. Gain insights into Kueue's real-world application in production self-managed clusters, serving machine-learning researchers, MLOps engineers, and data scientists. Understand how Kueue integrates with popular frameworks like DeepSpeed, PyTorch, Kubernetes Job, RayJob, and Jupyter to provide fair resource use and efficient management of accelerators and other resources.