Policy-based Compute Orchestration for AI/ML Workloads

Learn how to build robust AI/ML platforms on Kubernetes through policy-based compute orchestration in this 32-minute conference talk from KubeCon + CloudNativeCon. Discover how Outerbounds leveraged Kyverno to create an intelligent, extensible policy-based layer for orchestrating AI/ML compute workloads powered by Metaflow, addressing the complexities that Data Scientists face when managing workloads from local machines to complex cloud environments with diverse hardware and specialized GPU requirements. Explore practical strategies for using Kyverno's Cloud Native Policy-as-Code solution to dynamically route AI/ML workloads based on cost and resource availability, optimize resource scheduling for GPU-intensive tasks, and enforce security postures and compliance in multi-tenant Kubernetes environments. Gain insights into how policy-based orchestration can address challenges related to cloud providers, diverse hardware configurations, workload profiles, SLAs, cost pressures, and compliance requirements. See concrete examples of how Kyverno simplifies AI/ML platform operations while making them secure, compliant, and cost-effective for organizations building machine learning infrastructure at scale.