Managed Kubernetes Service achieves CNCF AI Conformance

We are proud to announce that UpCloud’s Managed Kubernetes Service (UKS) has officially achieved CNCF Certified Kubernetes AI Conformance for Kubernetes version 1.35.

Why This Matters

This certification provides independent, community-defined validation that our v1.35 clusters natively support the essential primitives required to run complex, stateful machine learning and artificial intelligence workloads reliably at scale.

Whether your team is deploying distributed training jobs, serving LLM inference models, or building long-running agentic workflows, this conformance guarantees interoperability and performance for:

• Dynamic Resource Allocation (DRA): Fine-grained hardware control over multi-node GPU clusters.
• Advanced Batch Scheduling: Native support for all-or-nothing scheduling frameworks (like Kueue) to maximize expensive accelerator uptime.
• Intelligent Autoscaling: Seamless Horizontal Pod Autoscaling (HPA) integration mapped to custom AI/GPU hardware metrics.

There is no separate configuration or add-on required. All existing and newly deployed Kubernetes v1.35 clusters are fully compliant with this new cloud-native standard.

To learn more about the testing framework and portability baselines, visit the official CNCF AI Conformance Program.