About
About
I’m an independent full-stack and systems engineer. For roughly 17 months I was the sole engineer on a production algorithmic trading platform — application through hardware, quant signals through customer support, a scope that’s normally a team. I think in systems: most of what looks like many separate problems is one underlying structure queried differently, and most failures live at the seams between parts rather than inside them.
How I work
The common thread across everything here — a local AI substrate, a game engine, a broken translation pipeline — is one instinct: reduce many apparent systems to a single primitive, and look for failures at the interfaces, not the components. I work review-first — read the system, form a plan, then build — because the expensive mistakes come from jumping straight to the fix. And I design around real constraints instead of wishing them away; several of the pieces here exist because a hard limit — a dead GPU, an 8K context window, a genre I don’t play — forced a better structure.
Background
Before engineering was the whole job, I worked where being wrong has consequences: front-line technical support at scale, and emergency medicine — ER/CNA and Emergency Medical Dispatch. Different domain, same discipline — make the call with incomplete information, under time pressure, and own the outcome. That is where the reliability habits come from: verify state before acting, keep what you know separate from what you’re inferring, and never let a confident guess stand in for a checked fact.
Contact
josh@colemanok.com
Medford, Oklahoma — remote.