AI Homelab and Self Hosted Web Tooling
A self-hosted cluster serving Mixture-of-Experts models beyond their rated VRAM and independent web services

I refined a method to pool GPUs from different vendors and generations into a single inference backend of roughly 20 GB over Vulkan. FreeToken, a bleeding-edge MoE inference server, fronts the cluster and handles the elastic inference, running Mixture-of-Experts models with far more total parameters than the cluster's VRAM would normally allow. My own scripts handle KV caching and system-prompt compaction, which is what makes the setup harness-capable.
The same rig trains as well as it serves, having refined Qwen SLMs and diffusion image models. All this is hosted alongside my personal web assets, like this website. I voluntarily subject myself to the maintenance of production-grade systems on every level of the tech stack to ensure that I survive the Big Tech apocalypse.
Built with FreeToken, ComfyUI, PEFT / TRL, Docker, Cloudflare, Reverse Proxy









