Jeff Geiser is VP, Customer Engineering at Zenlayer, focused on enterprise AI and sovereign inference infrastructure, based in Northern Virginia.
He helps large organizations build training and inference infrastructure, particularly at the edge. The conviction behind the rest of his work is simple: inference infrastructure people run themselves deserves the same rigor cloud infrastructure has had for a decade.
That splits into two problems he builds on. First, measuring and routing a sovereign stack: Wicklee gives real cost per million tokens and tokens-per-watt across your own GPUs.. how a system decides where inference should run, local or frontier, given what each node can actually do right now — is one I write about as the decision loop. Second, proving a sovereign stack is actually sovereign: bordercheck tests whether your data stays inside the border you drew by stressing the real stack and inspecting everywhere data lands.. no leaks!
Also: ongoing research into small private enterprise models (a fine-tuned Qwen2.5-7B built to run on owned hardware), writing on sovereign inference infrastructure, and base-building for running.