Coursera Flash Sale
40% Off Coursera Plus for 3 Months!
Grab it
Explore the platform-as-a-service layer that transforms GPU infrastructure into business-ready AI services through this 21-minute conference presentation. Learn how Mirantis k0rdent AI delivers composable service templates that enable operators to expose training, inference, and data services to tenants without requiring custom integration work for each new workload. Discover the catalog approach that allows operators to integrate frameworks like NVIDIA's Run.ai, Kubeflow, or MLflow based on customer preferences, with automatic provisioning of dependencies including cert-manager, GPU operators, and Argo. Watch a live demonstration of deploying an inference cluster template with GPU node profiles and Run.ai services, showcasing how complex manual installations are reduced to a single cluster deployment taking approximately fifteen minutes on AWS. Understand how the system optimizes cost efficiency through intelligent GPU node labeling and workload placement, ensuring GPU-backed nodes handle GPU workloads while CPU tasks run on appropriate nodes. Examine how enterprises can leverage this catalog internally for development acceleration or publish services externally for customers, with support for sovereign operators maintaining air-gapped local GPU hardware and hybrid operators extending across public and private GPU environments. Gain insights into delivering complex AI services quickly while maintaining governance and observability, eliminating the fragile manual integration typically required in GPU PaaS environments.
