I’m Jeff Geiser — VP, Customer Engineering at Zenlayer, building in public at the intersection of enterprise AI and sovereign inference. Based in Northern Virginia.
Day job: helping large companies deploy training and inference infrastructure, focused mostly on edge compute. Side project turned main bet: Wicklee, hardware-first observability and cost governance for self-hosted inference — the MPG rating for local AI. It measures what your own GPUs actually cost, down to $/1M tokens, instead of asking you to trust a vendor’s number.
The thread underneath both: inference infrastructure people run themselves deserves the same rigor cloud infrastructure has had for a decade — you should be able to see what it costs, where it’s running, and why. Wicklee is the observability piece. Increasingly I’m also writing about the layer above it: how a distributed system — one agent, or a whole network — decides where inference should run, and what it would take to name that as its own control-plane problem. That’s in the writing, framed as architecture thinking, not a product I’m building or claiming to own.
On the side I still run a small fine-tuning experiment — a 7B enterprise model, personal research, not a product.
When I’m not in front of a screen I’m running — base building right now, no race on the calendar yet.