Skip to content
WEKA

WEKA & Rafay - Architecting the Right Foundation for AI

WEKA and Rafay architect a GPU-cloud foundation that is more than stacked compute. Providers get performance, efficiency, and simplicity for multi-tenant AI clouds at exabyte scale.

Introduction

Building GPU accelerated clouds is no longer just about stacking compute. To succeed, providers need to deliver performance, efficiency, and simplicity in one cohesive platform.

AI Clouds, GPU-as-a-Service (GPUaaS) providers, sovereign AI initiatives, and private AI clouds face mounting pressure to maximize infrastructure density while managing complex multi-tenant environments at exabyte scale. Infrastructure teams struggle with power and space constraints that limit density, complex orchestration across bare metal and Kubernetes, and the challenge of delivering secure multi-tenant governance. Traditional approaches force impossible trade-offs between performance, cost, and operational complexity.

This paper shows how NeuralMesh by WEKA and the Rafay Platform together deliver the foundation needed to build efficient, automated AI clouds.

Challenges

Solution

Benefits

Storage Built for AI

NeuralMesh is designed from the ground up for AI workloads, validated through the NVIDIA Cloud Partner (NCP) Reference Architecture, and trusted by leading AI Clouds and sovereign AI Clouds.

Key Capabilities of NeuralMesh

Orchestration at Scale

The Rafay Platform provides the orchestration and governance layer that makes GPUaaS and private AI clouds consumable at scale.

Key Capabilities of the Rafay Platform

With Rafay, operators gain predictability and consistency across environments, whether public AI Clouds, private AI deployments, or sovereign platforms.

WEKA and Rafay: Purpose-Built for AI Clouds

AI Clouds choose WEKA and Rafay when they need:

Architecture diagram of the Rafay Operator orchestrating NeuralMesh by WEKA: infrastructure and storage admins, tenant onboarding and Day-2 blueprints, and per-tenant filesystems mounted to GPU compute via POSIX or CSI.

Integration that Streamlines Operations

The Rafay Platform orchestrates NeuralMesh via scoped API tokens. Tenant onboarding creates a dedicated Organization per Tenant, applies quotas/policies, and provisions per-tenant filesystems; apps consume storage via CSI (Kubernetes) or direct POSIX mounts on bare metal. Day-2 changes are executed as Rafay blueprints/workflows.

What’s Automated

Typical Onboarding Workflow: When a new ML team onboards, Rafay automatically creates a dedicated Organization with scoped credentials, applies tenant-specific quotas and security policies, provisions filesystems with requested capacity, and publishes storage to the team’s cluster—resulting in a ready-to-use environment in minutes.

Reference: WEKA Operator deployments https://docs.weka.io/kubernetes/weka-operator-deployments

Conclusion

By combining NeuralMesh by WEKA with the Rafay Platform, providers can deliver GPU-accelerated clouds that are:

For AI Clouds, GPUaaS providers, sovereign AI platforms, and enterprises, this means faster time-to-market, lower operational costs, and sustainable differentiation in the rapidly evolving AI cloud economy.

To learn more about Rafay, visit: https://rafay.co/

To learn more about NeuralMesh by WEKA visit: https://www.weka.io/

Did this page meet your expectations?

Scale Production AI Faster with NeuralMesh

Your models aren't slow. Your data is. Fix AI bottlenecks with high-throughput infrastructure.