AI enthusiast · Practical systems builder · Local-agent workflow explorer
AI 狂热追逐者 · 实用系统构建者 · 本地 Agent 工作流探索者
I am deeply fascinated by the current AI wave, but I care less about demos and more about systems that actually work.
我非常关注 AI 的快速演进，但我更关心的不是炫技 demo，而是 真正能落地、能验证、能长期运行的系统。
My current focus:
- Building practical AI workflows around local models, DGX, Mac mini, and agent tooling
- Turning messy real-world requirements into usable product systems
- Designing quality gates for AI-generated work, so AI agents stop shipping half-finished features
- Exploring domain-specific audit systems for insurance, healthcare data, and structured knowledge bases
我现在主要关注：
- 本地模型、DGX、Mac mini 和 Agent 工具链
- 把真实复杂需求做成可用产品系统
- 给 AI 交付建立验收和监工机制，减少半成品和假功能
- 探索保险、医保数据、行业知识库等垂直领域的审计系统
I am building an enterprise AI implementation framework that helps organizations turn business workflows into agent collaboration, knowledge systems, automation tools, and verification governance, so AI can evolve from isolated tools into sustainable productivity systems.
我致力于构建面向企业的 AI 落地体系，帮助组织将业务流程转化为智能体协作、知识库、自动化工具和验证治理机制，让 AI 从单点工具升级为可持续运行的生产力系统。
This is one of my current directions, alongside local AI infrastructure, AI delivery governance, and domain-specific audit systems.
这是我当前方向之一，同时我也在持续探索本地 AI 基础设施、AI 交付治理和行业审计系统。
A practical framework for building verifiable, governable, handoff-ready AI delivery systems.
AI 交付操作系统：一套让 AI 从可验证、可治理、可交接、可持续交付的方法论。
Core layers:
- Anti-bullshit delivery protocol
- Three-layer LLM-Wiki knowledge system
- Agent learning & handoff pipeline
- Enterprise implementation patterns
An open framework for turning business workflows into agent collaboration, knowledge systems, automation tools, and verification governance.
一个面向企业 AI 落地的开源实施框架，帮助把业务流程转化为智能体协作、知识库、自动化工具和验证治理机制。
Includes:
- AI readiness assessment
- workflow-to-agent mapping
- verification governance checklist
- implementation handoff template
- 90-day AI pilot playbook
A local quality gate for AI-generated product and code deliveries.
一个本地运行的 AI 交付监工工具，用于检查 AI 的交付说明、交接内容和代码变更是否真的可验收。
It catches:
- fake UI / mock data risk
- missing verification evidence
- missing acceptance criteria
- unauthorized phasing
- weak handoffs
- missing rollback or scope notes
A sanitized Hong Kong insurance product audit and presentation platform with a public site, admin workflow, and FastAPI backend.
一个已清理敏感信息的 香港保险产品信息审计与展示平台，包含公开前台、管理后台和 FastAPI 后端。
flowchart LR
A["Real-world problem<br/>真实问题"] --> B["AI-assisted workflow<br/>AI 辅助流程"]
B --> C["Local / private infrastructure<br/>本地与私有基础设施"]
C --> D["Product system<br/>产品系统"]
D --> E["Verification & handoff<br/>验证与交接"]
I like projects that combine:
- AI agents
- local LLMs
- product thinking
- data systems
- operational workflows
- verification and audit
我喜欢把 AI、产品、数据、流程和验证体系串起来，而不是只做一个看起来能跑的 demo。
No evidence, no delivery.
没有证据，不算交付。
I believe future AI systems need more than prompts. They need:
- context discipline
- verification evidence
- rollback plans
- handoff quality
- human-readable operating rules
- practical workflows that survive real use
我相信未来真正有用的 AI 系统，不只靠提示词，还需要上下文纪律、验证证据、回滚方案、交接质量和可持续运行的工作流。
I am building a personal open-source shelf around:
- AI Delivery OS — verifiable, governable, handoff-ready AI delivery methodology
- enterprise AI implementation frameworks
- AI delivery governance & verification
- local AI infrastructure
- domain-specific audit systems
- practical agent workflows
- bilingual open-source documentation
我正在搭建自己的开源作品架：AI 交付操作系统、企业 AI 落地框架、AI 交付治理与验证、本地 AI 基础设施、行业审计系统、实用 Agent 工作流，以及中英文开源文档。