DockerOPNsense • Advanced •
Private AI Stack: Ollama, Open WebUI, OPNsense, and Terraform
Reference architecture for private LLM workloads. Includes Docker Compose, OPNsense VLAN isolation, and Terraform provisioning.
Read Architecture →Sovereign AI Infrastructure for Modern Engineering Teams.
Deep technical guides on deploying private LLMs, securing AI workloads, and automating cloud and on-prem infrastructure with Docker, Kubernetes, OPNsense, and Terraform.
sys_topology.yml Operational
Ingress
User / Team
Tunnel
Cloudflare Tunnel
VLAN 50
OPNsense FW
Compute
Docker / K3s Host
vLLM / Ollama
Vector DB
Prometheus Logs
Featured Architecture
View all →
Hardware Independence
Deploy models on consumer GPUs or enterprise clusters without vendor lock-in. Full VRAM utilization guides.
Zero-Trust Boundaries
Isolate inference nodes via strict VLANs and deep packet inspection using OPNsense and Tailscale.
Data Sovereignty
Keep proprietary code and corporate documents off public APIs. Full control over the RAG pipeline.
Technical Playbooks
Browse full vault →Playbook Title
Implementation Focus
Est. Time
Status
Private AI Stack: Ollama, Open WebUI, OPNsense, and Terraform
OPNsense Security: High
Reference architecture for private LLM workloads. Includes Docker Compose, OPNsense VLAN isolation, and Terraform provisioning.
Time:2 Hours
Stable
Securing a Private AI Server with OPNsense VLANs
OPNsense Security: High
Step-by-step blueprint for isolating LLM inference workloads using OPNsense VLAN micro-segmentation and strict egress blocking.
Time:3 Hours
Stable
Terraform Module for RunPod GPU Provisioning
Terraform Security: High
Provision ephemeral GPU instances on RunPod using Terraform. Includes variable schema, outputs, and destroy lifecycle management.
Time:45 Minutes
Stable
Latest Deep Dives
AMD •
AMD ROCm GPU Passthrough under Docker with Linux cgroups v2
Step-by-step technical guide to exposing AMD GPU devices (/dev/kfd and /dev/dri) to Docker containers using ROCm.
VPS •
Best VPS Providers for Self-Hosted AI Workloads (2026)
Technical comparison of VPS and GPU cloud providers for self-hosted AI. Evaluated for VRAM, bandwidth, egress cost, and deployment constraints.
Security •
Multi-Tenant Container GPU Security and Hardening
Hardening containerized GPU access, preventing VRAM leaks, restricting privileged access, and securing host CUDA drivers.
Build your sovereign stack.
Join 5,000+ engineers receiving our weekly deep dives on securing, scaling, and automating private AI workloads.
Architectures referenced by engineers at enterprise security firms, financial institutions, and specialized SaaS providers.