Build Your AI Factory on Bare Metal
GPU sharing, ML pipelines, LLM serving — all on Kubernetes, on your hardware.
What You Get Today
GPU Passthrough
Full GPU passthrough to VMs and pods via KubeVirt for maximum performance.
Multi-Tenant GPU Isolation
Each project gets its own GPU resources with RBAC control — no cross-tenant interference.
KubeFlow Integration
ML pipeline orchestration with KubeFlow for training, tuning, and serving workflows.
GPU Monitoring
Prometheus metrics + Loki logging for real-time monitoring of GPU workloads.
VM-Based GPU Instances
GPU-enabled VMs with web console and SSH terminal directly from browser.
Coming in Q2 2026
- NowGPU Passthrough to VMs and pods via KubeVirt
- NowMulti-tenant GPU isolation with RBAC control
- NowKubeFlow integration for ML pipeline orchestration
- NowPrometheus metrics + Loki logging for GPU workloads
- SoonHAMI Integration — Fractional GPU sharing (one GPU → multiple tenants)
- SoonvGPU Support — Virtual GPUs for true multi-tenant GPU cloud
- SoonLLM Serving — Model inference infrastructure out-of-the-box
- SoonVector Databases — AI-native data stores
One Platform for Everything
Kube-DC is built on top of Kubernetes, so GPU workloads, containers, and VMs all live in one control plane. There are no separate systems for AI and cloud — one platform, one billing system, one access management through Keycloak. This unified approach reduces operational complexity and eliminates the need for separate GPU management tools.
Who Uses AI Factory
AI Training Clusters
Startups and enterprises running large-scale model training on dedicated GPU infrastructure.
GPU-as-a-Service
Cloud providers offering on-demand GPU instances — the highest margin revenue stream.
Private LLM Inference
Enterprises with data privacy requirements running LLM inference on-premises.
ML Pipeline Execution
Data science teams running end-to-end ML pipelines through KubeFlow on dedicated hardware.