Open-source software for running AI you own.
Tools to run models on your own hardware, know what they cost, and trust what you are running. Apache-2.0, built in the open, from a small independent company in Gig Harbor, Washington.
Projects
all Apache-2.0LLMKube
Run LLM inference on Kubernetes, across NVIDIA, AMD and Apple Silicon. A project I created and lead, built with a growing community of contributors.
Socair
Signed, fileable assurance reports for open-weight models, before they reach production.
InferCost
Know what every inference request costs on your own hardware, from GPU amortization to electricity.
Also from Defilan Technologies: MyGridCheck, power outage tracking for the Puget Sound region →
Writing
all posts →- KubeCon, a Raspberry Pi, and five first-time contributors llmkube.com
- I asked a self-hosted DeepSeek to build a plugin for its own harness llmkube.com
- My Kubernetes operator passed every health check while it ignored every change llmkube.com
- Trying a new engine on the Strix Halo: 41 minutes to first token became 3.5 llmkube.com
About
Contact
For collaboration, sponsorship, or support for any of the projects: contact@defilan.com