Quiet systems.
Loud results.
I build self-hosted AI, run a 22-node home cluster that thinks like one machine, and ship product end to end - from kernel to checkout.
- 14
- 136
- 1
A few layers down from the dashboard.
Self-hosted AI
Fleet operations
Shipped product
Agent doctrine
Observability
Hardening & ops
Five live reads, backed by public signals.
The headline is backed by real numbers: fleet health, deployed properties, current work, and backup state - the same public-safe data that powers the rest of the site.
S.A.M - one mind, many machines.
Twenty-two pieces of hardware - workstations, servers, laptops, GPU boxes - running a shared doctrine. One operator, one source of truth, and just enough automation that the machines stay in agreement when I'm not looking.
Open fleet evidenceThis one shipped from a text thread.
Poke is the iMessage-native agent that designs and deploys these sites. I describe what I want; it builds, commits, and ships — then the same build runs in two homes: Poke’s cloud and my own fleet.
Describe it over text
Ship to one repo
Deploy to both
Everything ships from one desk.
Every property runs on the same fleet, shares the same auth and edge layer, and ships from the same operator desk.
If the work looks like your problem, get in touch.
I take on a small number of consulting engagements when the fit is right - infrastructure design, AI ops, or full-stack delivery that needs a senior pair of hands.
Start a conversation





