Workshops Archive

Simplifying Compute Access for RL Use Cases at Scale


Workshop: 8th International Workshop on Containers and New Orchestration Paradigms for Isolated Environments in HPC (CANOPIE-HPC)

Authors: Brandon Jacobs (CoreWeave, Inc.)

Abstract: As the demand grows for larger reinforcement learning (RL) rollouts, evaluation tasks, and tighter training loops, scaling your compute beyond traditional Kubernetes cluster boundaries becomes essential. CoreWeave Sandboxes provide a single API interface and programmable execution environment that aggregates capacity across your own Kubernetes clusters and serverless infrastructure. This allows you to run parallel RL environments, scale evaluation tasks, and burst compute seamlessly without changing your API or SDK behavior, regardless of where your capacity physically exists. By maintaining a consistent workflow across distinct compute locations, researchers can focus on accelerating RL experiments rather than managing infrastructure limits.
This talk explores how to unify compute across disparate clusters and serverless capacity through a single API, enabling seamless scaling for heavy RL rollouts and parallel evaluations without modifying your existing workflow.


Back to 8th International Workshop on Containers and New Orchestration Paradigms for Isolated Environments in HPC (CANOPIE-HPC) Archive Listing

Back to Full Workshop Archive Listing