NVIDIA/aicr

What it solves

Running GPU-accelerated Kubernetes clusters is often unstable due to complex dependencies between kernel versions, drivers, container runtimes, and Kubernetes releases. AI Cluster Runtime (AICR) eliminates this guesswork by providing validated, version-locked configuration "recipes" that ensure reproducible and optimized deployments across various hardware and cloud environments.

How it works

AICR uses a recipe engine that matches a user's target environment (cloud, GPU, OS, and workload intent) against a library of validated overlays and mixins. The system then renders these recipes into deployment-ready bundles for popular GitOps tools like Helm, Argo CD, Flux, and Helmfile. To ensure the cluster is running as intended, a Snapshot Agent captures the live state of the cluster, which the AICR validator then compares against the recipe to detect configuration drift or performance issues.

Who it’s for

It is designed for platform engineers and DevOps teams managing GPU-accelerated Kubernetes clusters for AI training and inference workloads across multiple cloud providers (e.g., AWS, GCP, Azure) or on-premises hardware.

Highlights

• Multi-Deployer Support: Generates artifacts for Helm, Argo CD, Flux, and Helmfile.
• Automated Validation: Includes deployment, performance, and conformance phases to verify cluster health.
• Supply Chain Security: Implements SLSA Build Level 3 provenance, signed SBOMs, and image attestations.
• Broad Compatibility: Supports a wide range of NVIDIA GPUs (H100, A100, B200, etc.) and various operating systems (Ubuntu, RHEL, Amazon Linux).
• Drift Detection: Ability to compare cluster snapshots to identify configuration changes over time.