Homelab v3 mostly for local AI and development
retained text
Been building and upgrading my homelab for a while now. This is the current setup, mostly used for local AI, development, experimenting with infrastructure and self hosting random stuff **networking** Running Ubiquiti for pretty much everything. Being able to isolate IoT devices, smart TVs, random Wi-Fi devices and local AI sandboxes has been really useful I also have 10Gb internet from my ISP with 4G failover and a UPS 2U Pro to keep everything running **proxmox server** Main server running Proxmox with the following hardware: * **CPU:** AMD Ryzen 9 5950X * **RAM:** 128GB Micron ECC DDR4 * **GPUs:** 2 GPUs, one Intel Arc for Jellyfin transcoding and another 4090 for gaming/CUDA experiments Running my media server with ZFS and Jellyfin, Home Assistant, development LXCs with GUI desktops for myself and AI agents and a bunch of other services **local AI** Dell Pro Max GB10 mainly for running local LLMs like Qwen 3.8 27B with lots of small concurrent requests Also running STT, TTS, embeddings, image generation and OCR models in separate Docker Compose setups Besides actually using the models, I wanted something to experiment with CUDA and replicate production like environments for work **mac minis and mobile devices** 2x M4 Mac minis for AI agents that need macOS or iOS development environments without paying cloud providers At the bottom of the rack there's also another MacBook Pro along with iPhones and Android devices connected to my AI agents for mobile development and testing **k3s cluster** A small cluster of HP EliteDesks that I use to experiment with Kubernetes configurations Currently also experimenting with giving each LLM chat its own ephemeral sandbox in the cluster so agents can execute code and work in isolated environments
document fetched 2026-11-20 00:37 · canonical https://reddit.com/r/homelab/comments/1y0g201/homelab_v3_mostly_for_local_ai_and_development