About Project

Neuronix Tech Private Cloud

Smart Fusion built Neuronix Tech's in-house private cloud on Proxmox VE, a four-node cluster on HP ProLiant servers with two dedicated GPU workstations for AI training and inference, plus the storage, networking, high availability, backup, and monitoring to run it in production.

About Client

Neuronix Tech

Neuronix Tech was running its products across standalone servers and renting cloud GPUs by the hour for AI work. It wanted one platform it owned and controlled: predictable costs, data kept in-house, headroom to grow, and dedicated GPU compute always on hand. Smart Fusion designed and deployed a private cloud on a Proxmox VE cluster of four HP ProLiant servers (two DL380 and two DL360), migrating the company's platforms, from e-commerce and HRMS to the Neuro-ID service and the proxy gateway, into isolated virtual machines, and adding two Dell Precision 7920 GPU workstations as dedicated AI machines for training, fine-tuning, and inference. The engagement covered cluster setup, shared storage and snapshots, network segmentation, GPU passthrough, high availability, automated backups with tested restores, monitoring, and a full administrator handover.

View site

Every Neuronix Tech platform running as an isolated VM on a four-node Proxmox VE cluster

A four-node Proxmox VE cluster

Standalone servers were replaced by a Proxmox VE cluster where every product runs as an isolated virtual machine, live-migratable between nodes and administered from one console.

Four-node cluster on HP ProLiant DL380 and DL360 servers, managed as a single platform
The company's platforms migrated into isolated KVM virtual machines and LXC containers
Shared storage with instant snapshots, thin provisioning, and scheduled replication
High availability, so critical VMs restart automatically on a healthy node if one fails

Proxmox VE datacenter summary showing cluster health, running virtual machines and containers, and resource usage

Cluster health, guests, and resource usage for the whole platform in a single view

Two GPU workstations for AI

Two dual-socket workstations were added as dedicated AI machines, with full GPU passthrough so models train and serve at native hardware speed.

Two Dell Precision 7920 workstations attached to the cluster as dedicated AI nodes
Full PCIe GPU passthrough (IOMMU / VFIO), with no virtualization overhead on the accelerator
NVIDIA drivers, CUDA, and a container runtime pre-configured for training and inference
GPU utilization, memory, and temperature monitored alongside the rest of the platform

Proxmox VE node summary with CPU, memory, and hardware detail plus a live usage graph

Per-node compute, memory, and live utilization, including the GPU workstations

Networking, backup, and operational control

The platform was hardened and made operable from day one, with segmented networking, off-cluster backups, monitoring, and documentation the Neuronix team owns.

VLAN segmentation and firewall rules separating management, cluster, workload, and AI traffic
Secure administration through the proxy gateway and VPN, with no management interface exposed to the internet
Scheduled cluster-wide backups to a dedicated target, with periodic tested restores
Metrics dashboards, alerting, runbooks, and an administrator handover session

Proxmox VE storage summary showing backup content type and capacity usage over time

Storage backing VM disks, ISO images, templates, and scheduled backups

The outcome

Neuronix Tech now runs its platforms and its AI workloads on a Proxmox VE cluster it fully controls. Four HP ProLiant nodes give it headroom and high availability, two GPU workstations are available on demand for training and inference, backups run and are verified on a schedule, and Smart Fusion delivered the documentation and training for the Neuronix team to operate and expand the cloud on its own.

Technology stack

Proxmox VEProxmox VE Cluster & HAKVM / QEMULXCHP ProLiant DL380 / DL360Dell Precision 7920 (GPU workstations)ZFSGPU passthrough (IOMMU / VFIO)NVIDIA CUDADockerUbuntu Linux

Delivered through Cloud Engineering & Implementation.

View the Service