I'm Liang Cheng (梁程 / Ray), an AI-native product engineer based in Nanjing. I turn fuzzy ideas into working, testable software—from tool-using agents and local-first RAG systems to polished web and iOS products.
NOWBuilding real agent systems, reusable AI workflows, and stronger backend foundationsBIASTool use over wrapper demos · evidence over pitch · shipping over slide decksPASTBuilt telecom-grade React, TypeScript, and ECharts systems at HuaweiEDUQUT Computer Science graduate · IELTS 7.5
const liang = {
focus: ["agent systems", "full-stack products", "indie iOS"],
aiStack: ["Codex", "Kimi", "Qoder"],
buildLoop: ["understand", "prototype", "verify", "ship"],
principle: "Build the proof, not just the pitch.",
} as const;My daily AI build stack: Codex for agentic coding, Kimi as a reasoning partner, and Qoder as an AI coding workspace.
Three anonymized challenges from real hiring processes—and my production-minded solutions—covering a Computer Use agent dashboard, a React Native screenshot-action agent, and a Jira-style task system. The repository also distills a reusable playbook for scope, architecture, testing, evidence, and delivery.
|
|
|
A stateful 3D-building agent. Models choose typed tools; the host validates every mutation and renders the result in a Three.js editor.
|
🔬 SciFlowA local-first research workspace with a React frontend, FastAPI backend, PostgreSQL/pgvector, and privacy-minded document retrieval.
|
|
A safety-first Agent Skill for diagnosing macOS storage pressure with read-only scans, reversible quarantine, and explicit purge gates.
|
A growing, bilingual library of production-grade AI Agent Skills—portable workflows with clear boundaries and observable outcomes.
|
★ Stars earned across my public repositories · refreshed every 6 hours
Portfolio · Writing · All repositories · Let's build something
Build the proof. Keep the taste. Ship the thing.