Sovereign AI Infrastructure 101 — Vancouver

Sovereign AI is no longer a research-lab abstraction. Operators in Vancouver are spinning up GPU clusters in their own colocation cabinets, running 70B-class models against private datasets, and pulling workloads back from hyperscaler APIs that quietly meter every token. This evening session is for the people doing that work, or about to.

We will walk through a working reference stack: dual-H100 host, NVMe scratch, vLLM serving a quantized 70B, and a small inference proxy that keeps audit logs local. No slides full of vendor logos. Real config files, real cost math, and the specific failure modes you will see in the first thirty days. The 90-minute talk maps the territory; the 30 minutes after is unstructured Q&A so you can press on whatever is breaking in your own setup.

Free. RSVP-only. Fifteen seats. This is intentionally small so the conversation stays specific. If you are evaluating sovereign infrastructure for a client, a startup, or an internal team, you are the audience. Confirmed attendees receive the venue address and a short pre-read forty-eight hours before the date.