Homelab

docs.bhamm-lab.com · github.com/blake-hamm/bhamm-lab

A homelab built for AI/ML experimentation, DevOps exploration, and self-hosting — documented as infrastructure-as-code in a monorepo.

Hardware: 6 servers across SuperMicro, Framework mainboards, and custom builds, with a Ceph storage backbone, dual 10GbE networking, and accelerated compute (AMD Radeon AI Pro R9700, Intel Arc A310, dual AMD Ryzen AI MAX+ 395).

Software Stack: Ephemeral Talos Kubernetes clusters managed via ArgoCD + OpenTofu provisioning. Debian/Proxmox hosts configured with Ansible. NixOS for workstations. Ceph RGW + Garage for storage.

AI/ML: Llama.cpp inference scaled to zero with KubeElasti, routed through LiteLLM. GPU orchestration via Intel GPU plugin and AMD ROCm operator. Embedding inference on AMD Radeon, dense/MoE inference on dual Strix Halo systems.

Key Practices: GitOps deployments, 3-2-1 backup strategy with Argo Workflows + Backblaze B2 offsite (Cloudflare R2 standby), SOPS + Vault for secrets, network segmentation with OPNsense + Pi-hole DNS, automated disaster recovery with blue/green cluster testing.

The full docs site covers architecture , deployments , operations , AI/ML , backups , and security .