8000 stephen-taipei (Stephen Chuang 莊宗憲) · GitHub
[go: up one dir, main page]

Skip to content
View stephen-taipei's full-sized avatar
🎯
Focusing
🎯
Focusing

Highlights

  • Pro

Block or report stephen-taipei

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Content in all repositories owned by your account will be closed.
Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
stephen-taipei/README.md

Stephen Chuang

English · 繁體中文 · 简体中文 · 日本語 · 한국어 · ไทย

莊宗憲 · AI-Native Systems Builder · Agentic Systems Engineer · Tech Lead

I turn ambiguous problems into concrete systems, workflows, and production outcomes.

Building Better Workflows, an evidence-first reliability and delivery control layer for AI engineering agents.

My current focus is the boundary between probabilistic AI agents and deterministic software delivery.

AI can propose and reason. Production systems still need evidence, authority, validation, and proof of outcome.

13+ years of software engineering across architecture, full-stack systems, CI/CD, production delivery, engineering leadership, and AI-native development.

What I'm Building

Agentic Infrastructure

These tools govern AI coding agents from intent to real repo changes through evidence, authority, and validation.

  1. Better Workflows (flagship)

    Evidence-first AI engineering reliability + delivery gatekeeper

    betterworkflows.dev · Source · npm

    V6 Core 6.0.0-alpha.5 — free to install (Apache-2.0). Claude Code first; an experimental Codex adapter is included. V5 is frozen at RC1 (AGPL-3.0-only).

    npm install -g @better-workflows/core@alpha
    claude plugin marketplace add stephen-taipei/better-workflows
    claude plugin install better-workflows-core@better-workflows --scope local
    bw init

    Current evidence · At most once · Policy, not prompt · Fail-closed

  2. Deckhand: Multi-AI toolbar for Claude Code: usage gauges, model switching, parallel sub agents, read-only second opinions from Codex/Cursor/agy.

Also open source: Adorable Git Graph: Your git history as a cartoon you can scroll: every commit is a ball with a face, every branch a flowing pipe. Demo · npm

All open-source projects →

Shipped Product

Connectors (脈客) · Private repository

AI-powered personal CRM, shipped to production

ctrs.app · App Store

  • Role: Product direction, architecture, AI-assisted delivery orchestration, Code Review, QA, and iOS release delivery.
  • Status: Live on iOS and the web · Android in development
  • AI-assisted implementation context: React, NestJS, tRPC, SwiftUI, and Kotlin Compose.

Engineering Track Record

Thirteen years of production engineering is why the reliability work above is grounded in real delivery systems rather than demos. AI agents in my workflows touch Git, CI, deployment, databases, and real repositories.

Selected production architecture work from my frontend engineering background:

  • Designed and led an Angular/Nx architecture supporting 100+ themes, reducing a representative production theme leaf to 23 lines.
  • Reduced a theme CSS request from approximately 2.4 MB to 17 KB (~99%) and improved measured Lighthouse Performance from 50+ to 90+.
  • Directly led 3–4 frontend engineers and coordinated several support engineers who did not report to me; hosted occasional internal tech talks.

Scope, baseline, and what is explicitly not claimed for each result.

Core Expertise

  • Systems building: turning ambiguous problems into explicit models, boundaries, workflows, and executable systems
  • Agentic systems: AI coding agents, multi-agent workflows, agent orchestration, tool-driven execution
  • Agent reliability: evidence-bound validation, stale-state detection, reconciliation, fail-closed execution
  • Engineering governance: policy as code, deterministic validation, bounded automation, safe delivery
  • Software architecture: system boundaries, explicit ownership, behavior-preserving refactoring
  • Production engineering: Git, CI/CD, Linux, deployment, observability, databases
  • Engineering foundation: Node.js, NestJS, Angular, TypeScript, Nx, Laravel, PostgreSQL, MySQL, Redis

Connect

Selective conversations around agentic systems, AI agent reliability, developer infrastructure, and OSS collaboration are welcome.

Taiwan · Remote

Website · Articles · LinkedIn · Better Workflows · Connectors · Case Studies

Pinned Loading

  1. better-workflows better-workflows Public

    Stops AI coding agents from finishing on stale tests or repeating unseen side effects. Claude Code first.

    JavaScript

  2. deckhand deckhand Public

    Deckhand — the multi-AI toolbar for Claude Code: usage gauges, one-click model switching, parallel sub agents, and read-only second opinions from Codex, Cursor agent and agy.

    TypeScript

  3. adorable-git-graph adorable-git-graph Public

    Read-only, cartoon-style Git graph for GitHub and local repos — compare branches, find unique commits, check what changed.

    TypeScript

0