Notes from
the metal.
Engineering write-ups on how we build and run production systems — no content marketing, just what we actually do and what it costs.
Topics: bare-metal infrastructure and cluster economics, private AI and local LLM deployment, e-commerce and SaaS architecture, and the security work behind systems we operate for clients. Written by the people who run the hardware — every claim here is backed by a system we can point at.
New posts land as we ship — infrastructure notes, deployment walkthroughs, and the occasional post-mortem when something deserves one.
Why We Run Client Workloads on Bare Metal Instead of AWS
The cloud is a rental with a meter. We did the math on our own cluster — and the case for owning the metal is stronger than the brochure says.
read →Private LLM Deployment: Running AI Without Sending Data to OpenAI
Zero-egress inference isn't a paranoia feature — it's how you make AI a fixed cost and keep client data legally clean. Here's the actual architecture.
read →