Skip to main content

Kubernetes AI Conformance — QKS Evidence

This page publishes the self-assessment evidence for the QKS / kosmosai platform against CNCF Kubernetes AI Conformance. For every requirement, the exact commands executed and their output are included verbatim so that the submission's evidence field can reference them.

This evidence is submitted under ORKESTRIX, the QUANTUM C&S Kubernetes distribution (which includes QKS); the AI capabilities shown here run as components of the ORKESTRIX stack, many of them QKS-branded.

Program summary

CNCF AI Conformance verifies whether a Kubernetes platform has the foundational capabilities to reliably run AI/ML workloads (training, inference, agentic). It builds on the base Kubernetes Conformance certification and currently uses a self-assessment + public-evidence submission model.

:::note Checklist version This evidence targets the v1.36 checklist. dra_support (MUST in v1.35) was removed in v1.36 and is therefore out of scope. :::

Evidence status (MUST items · v1.36)

ItemCategoryStatusEvidence
ai_inferenceNetworkingMet (evidence included)View
gang_schedulingSchedulingMet (evidence included)View
cluster_autoscalingScheduling (conditional)Met (condition does not apply)View
pod_autoscalingScheduling (conditional)Met (evidence included)View
accelerator_metricsObservabilityMet (evidence included)View
ai_service_metricsObservabilityMet (evidence included)View
secure_accelerator_accessSecurityMet (evidence included)View
robust_controllerOperatorMet (evidence included)View

Evidence status (SHOULD items · optional)

ItemCategoryStatusEvidence
gpu_sharingAcceleratorsMet (evidence included)View
note

Each evidence page's code blocks contain the commands and output executed on a real cluster, verbatim. Reviewers can re-run the commands to verify the results.