Sheet 04 · General notes
About
I build systems that have to be right — and the tools that help other engineers do the same.
General notes
- 01
Background
I started in 2005 writing Linux C and TCP code for telecom back-ends in Dalian, then led a team of ten on systems for China Unicom. At Fidelity I led the first iOS investment app for the UK market. After moving to the US, I led the backend team at Fuhu, where we rebuilt user management on event sourcing for a million devices. At Oracle, over nine years, I built developer platforms and, most recently, LLM evaluation.
- 02
Correctness first
I care most about systems that are right, and stay right as they change. That means modeling the domain carefully, making illegal states unrepresentable, and letting the type system carry as much of the proof as it can. It led me through functional programming — PureScript, F#-style domain modeling, and today TypeScript with Effect.ts.
Notes on functional domain modeling ↗ - 03
Agents, carefully
LLMs and coding agents are now part of how I build. I prefer a thin harness, an explicit workflow, hard isolation and a person at the gates that matter — rather than clever autonomy. Lingtai is that idea written down as a tool.
Lingtai → - 04
Building alone, leading teams
I have led teams of six to ten, and I hold a PMP. I have also shipped whole products by myself: Helium and LLM-as-judge at Oracle were one-person efforts from idea to adoption, and so is nextloom.ai. Either way, I reach for practical tools over process — the best measure of a platform is the hours it gives back.
At a glance
Resume →- Role
- Principal Software Engineer · Founder, nextloom.ai
- Experience
- 20+ years, since 2005
- Based in
- Houston area, Texas
- steven.zhang031@gmail.com