Profile
I’m Luhui Dev, a developer researching and deconstructing agent engineering while exploring practical applications of AI in education. I focus on Agent Harness engineering, LLM application engineering, AI for Math, and education SaaS productization.
My current work connects two layers. The first is Agent engineering: context, tools, memory, permissions, runtime, observability, evaluation, and reliability. The second is domain application: using AI for Math and formal reasoning to build dependable education products.
I work by keeping product judgment and engineering execution in one loop, starting with small pilot scopes and then hardening what works into reusable systems. I care more about long-term reliability, collaboration, and iteration than short-lived trends.
Capability Matrix
| Domain | Focus |
|---|---|
| Agent | Harness engineering, runtime, context, evaluation, reliability |
| LLM | Application architecture, retrieval, orchestration, cost control |
| Math | Geometry runtime, constraint modeling, reasoning workflows |
| Product | Education SaaS, pilot design, roadmap sequencing, growth loops |
Why this direction
Agent systems become useful when they are reliable, observable, and able to operate within explicit boundaries. Math and education provide a demanding environment for testing those properties because intermediate reasoning and final outcomes can both be examined.