I build AI systems that fit into real engineering workflows.
我专注于构建能够真正嵌入研发流程的 AI 系统。

My focus is not just integrating models, but turning AI capabilities into systems that are:
我关注的不只是接入模型，而是把 AI 能力组织成具备以下特征的系统：

• deployable / 可部署
• configurable / 可配置
• measurable / 可度量
• policy-aware / 具备策略约束
• operationally reliable / 运行上可靠

I work at the intersection of:
我的工作主要位于以下交叉点：

• Applied AI / 应用型 AI
• Developer productivity / 研发效率
• Platform and tooling engineering / 平台与工具链工程
• Production-grade system design / 生产级系统设计

I’m especially interested in AI systems that improve how engineering teams work:
我尤其关注能够提升工程团队协作效率的 AI 系统：

• AI-assisted code review / AI 辅助代码审查
• policy-aware review workflows / 带策略约束的审查工作流
• feedback-driven engineering systems / 基于反馈闭环的工程系统
• engineering knowledge workflows / 工程知识工作流
• intelligent developer tools / 智能开发者工具

I care about the parts that make AI usable in practice:
我关注让 AI 真正可用的那些关键部分：

• workflow orchestration / 工作流编排
• context construction and grounding / 上下文构造与 grounding
• feedback loops / 反馈闭环
• evaluation and quality control / 评估与质量控制
• observability, configuration, and operations / 可观测性、配置与运维

My background is strongly shaped by platform and tooling work:
我的技术背景也深受平台与工具系统方向影响：

• developer tools / 开发者工具
• reverse proxy and infrastructure systems / 反向代理与基础设施系统
• configurable runtime platforms / 可配置运行时平台
• terminal and local developer experience / 终端与本地开发体验

AI-powered code review and engineering workflow system for Gitea
面向 Gitea 的 AI 代码审查与工程工作流系统

Highlights / 特点：

• staged review orchestration / 分阶段审查编排
• pluggable LLM providers / 可插拔 LLM Provider
• policy-aware behavior / 策略感知行为
• feedback loop foundations / 反馈闭环基础能力
• admin dashboard and runtime configuration / 后台管理与运行时配置
• developer-facing workflow integration / 面向研发流程的集成能力

High-performance reverse proxy platform built with Bun and TypeScript
基于 Bun 和 TypeScript 构建的高性能反向代理平台

Highlights / 特点：

• production-oriented systems design / 面向生产的系统设计
• configurable runtime behavior / 可配置运行时行为
• extensibility and plugin direction / 可扩展性与插件化方向
• developer-native infrastructure engineering / 面向开发者的基础设施工程

Modern Android terminal emulator focused on product-quality engineering and developer experience
聚焦产品级工程质量与开发体验的现代 Android 终端模拟器

Highlights / 特点：

• terminal workflow design / 终端工作流设计
• Kotlin / Compose product engineering / Kotlin / Compose 产品工程
• UX refinement for technical users / 面向技术用户的体验优化
• practical local tooling experience / 实用的本地工具体验

When building AI systems, I care about:
在构建 AI 系统时，我更重视：

• usefulness over novelty / 有用性优先于新奇性
• workflow fit over demo effect / 贴合工作流优先于 demo 效果
• grounded output over generic generation / grounded 输出优先于泛化生成
• measurable quality over vague intelligence claims / 可度量质量优先于模糊的“智能”描述
• product reliability over one-off prototypes / 产品可靠性优先于一次性原型

I’m continuing to deepen work in:
我正在持续深化以下方向：

• AI-native engineering workflows
• intelligent code review systems
• grounding and evaluation for developer-facing AI
• platform-grade AI applications for teams

对应中文：

• AI-native 工程工作流
• 智能代码审查系统
• 面向开发者 AI 的 grounding 与评估
• 面向团队的平台级 AI 应用

The long-term goal is to build AI systems that engineers can actually adopt in production workflows.
长期目标是构建能被工程团队真正纳入生产流程的 AI 系统。

A few principles behind my work:
我做系统时遵循的几个原则：

• AI should be part of a system, not the whole system.
  AI 应该是系统的一部分，而不是系统的全部。
• Good workflows matter as much as good models.
  好的工作流和好的模型同样重要。
• Configuration, policy, and feedback are first-class product features.
  配置、策略和反馈闭环是一级产品能力。
• Evaluation is a core engineering concern, not an afterthought.
  评估是核心工程问题，而不是事后补充。
• The best AI tools feel operationally reliable, not just impressive in a demo.
  最好的 AI 工具应当在运行上可靠，而不只是 demo 好看。

Applied AI engineer focused on developer productivity, intelligent engineering workflows, and production-grade AI systems — with strong foundations in platform, tooling, and systems design.

专注于研发效率、智能工程工作流与生产级 AI 系统的应用型 AI 工程师，具备扎实的平台、工具链与系统设计基础。