Juno Orion - GPU and Workload Orchestration for Amazon EKS User Seats

OVERVIEW Enterprise on-premises GPU utilization sits at 10-15%. The gap is not hardware. It is the absence of a scheduling layer that treats every GPU, CPU, VM, and container as one shared, policy-driven pool. Teams pay for idle capacity. Engineers manage provisioning queues instead of building.

Juno Orion is the unified compute plane for GPU/CPU-intensive workloads on AWS. Customer-hosted and production-hardened, Orion deploys into your existing VPC alongside Amazon EKS and turns your GPU fleet into a shared, request-driven resource pool. Containers, VMs, and bare metal run under one management layer. Provisioning takes 60 seconds. Juno is validated on NVIDIA DGX and H100 through NVIDIA Inception.

THE PROBLEM ORION SOLVES GPU cycles go to waste. Clusters sized for peak demand sit at 10-15% utilization the rest of the time. Provisioning creates friction. Standard EKS workload setup requires Kubernetes expertise, IT ticketing, and 24-48 hours per session. That overhead compounds across every engineer, researcher, and artist on your team. VMs and containers live in separate worlds. Different tools, different policies, no shared scheduler. Orion solves all three from a single management layer.

KEY CAPABILITIES

GPU-Aware Scheduling and Workload Bin-Packing Orion packs multiple concurrent workloads onto a single GPU node using time slicing with no contention. CPU, RAM, and GPU are treated as a unified pool allocated by policy. Teams typically see 2-4x more capacity from the same hardware footprint.

60-Second Self-Service Provisioning End users launch workstations and compute environments in 60 seconds from a self-service portal. No IT ticket. No Kubernetes knowledge required. Administrators define templates once; users click and go.

VMs and Containers on the Same Cluster KubeVirt runs full Windows VMs alongside Linux containers on the same EKS cluster with GPU passthrough included. No separate hypervisor contract. On AWS, Crossplane provisions EC2 instances directly from the Orion workload catalog. End users click once. Juno handles networking, security group assignment, authentication brokering, and lifecycle management with no public IP exposure.

Demand-Driven Autoscaling The cluster scales up when users request compute and terminates resources the moment sessions end, so capacity follows demand across every user on the cluster.

Security and Isolation Projects map to Kubernetes namespaces with enforced CPU, RAM, and GPU quotas per team. Network policies block east-west traffic between namespaces. Orion consumes JWT tokens from any identity provider including Okta, Google Workspace, Cognito, and Active Directory.

HOW IT WORKS ON AWS Orion deploys into your existing VPC alongside Amazon EKS and integrates with GPU-accelerated EC2 instance types including G5, P4, P3, and Inf2. All traffic routes through Juno's ingress and authentication layer with no public internet exposure by default. Orion also runs across on-premises and fully air-gapped environments from the same management layer. Your data never leaves your environment.

PRODUCTION PROOF Orion has been running under live GPU-intensive workloads since April 2025 with zero critical failures. R3D Studios, an AI-accelerated film production studio, runs Orion in production on Amazon EKS:

• Up to 40% AWS compute cost reduction (EC2 On-Demand, no credits applied)
• 2:1 GPU density: 10 concurrent users across 5 GPU-accelerated nodes
• 60-second workstation provisioning, down from 24-48 hours
• Infrastructure team reduced from 5 to 2 engineers, same production workload
• 24/7 global production with zero infrastructure tickets

"Juno just works for us. R3D wouldn't exist without Juno." - Donald Strubler, Head of Development and Co-Founder, R3D Studios

WHO RUNS ORION AI/ML and LLM teams running multi-tenant GPU clusters on EKS. VFX and render pipelines needing GPU workstations provisioned on demand. HPC and research computing teams sharing GPU and CPU pools across batch jobs and simulation pipelines. VDI environments requiring Windows and Linux workstations in 60 seconds with no public internet exposure.

BUILT FOR AWS Orion extends your existing AWS investment, running on EKS, provisioning EC2 resources, and integrating with your existing VPC, IAM, and storage configurations.

GET STARTED Subscribe through AWS Marketplace to get started. For volume pricing or custom terms, use Request private offer on this page.