我把 AI 从炫技 Demo 做成企业敢上线的系统。
I turn AI prototypes into systems teams can ship, audit, and trust.
中文
不止做聊天框。我做的是 AI 落地的控制层：数据从哪来、结果怎么评测、谁来审批、如何审计、怎样部署上线。
我是 Zack Wang，10+ 年 Java、TypeScript、AWS、企业系统和全栈工程经验。现在聚焦 Applied AI / Enterprise AI Delivery：把 LLM、评测、审批、安全边界、AWS 数据系统和产品工作流组合成真正可部署、可演示、可持续迭代的系统。
| 我能带来的能力 | 为什么重要 |
|---|---|
| 10+ 年企业软件工程经验 | 理解真实系统、发布压力、数据边界、线上支持和协作成本。 |
| Java / TypeScript / AWS / 数据工作流 | 能把 AI 功能接进企业已有技术栈，而不是停留在孤立 Demo。 |
| LLM 评测、可观测性、审批门禁 | 能设计可测试、可监控、可治理的 AI 系统。 |
| 产品化项目探索 | 不只做实验，而是持续把项目推向真实可用的业务流程。 |
| 项目 | 证明什么能力 | 链接 |
|---|---|---|
| JTestGen AI Java 单测生成与覆盖率修复 |
JaCoCo 定位、JUnit/Mockito 生成、Maven 校验、失败修复循环、CI 友好产物。 | Repo |
| DREAM 团队知识与工程自动化记忆平台 |
Open-core / private extension 架构，memory packs、codebase index、Evidence Graph、governed memory distillation、Requirement Case、PR review、scorecard、provenance/security 边界和 SQLite audit trail，展示 source-backed AI 工程工作流。 | Repo |
| Enterprise AI Engineering Orchestrator 发布风险、测试建议与人工审批工作流 |
用仓库上下文、风险评分、测试建议、UiPath Test Manager proof、DynamoDB proof 和 human approval gate 展示可审计的企业工程 AI 编排。 | Repo · Demo · Proof |
| LLM Eval Observability LLM 评测与可观测性 |
Provider adapters、质量门禁、SQLite run history、可重复 AI 应用检查。 | Repo |
| AI ContentOps Studio AI 内容生产工作流 |
研究、生成、质量评分、人工审核、发布证据和部署配置的可追踪内容系统。 | Repo |
| AI Commerce Copilot 跨境电商 AI 运营助手 |
选品评分、供应商询价、内容草稿、人工审批、审计证据、本地/私有部署边界。 | Portfolio |
| Brand Publishing Pipeline 技术品牌发布流水线 |
技术博客、短视频脚本、平台 metadata、审核门禁、安全发布交接。 | Repo |
企业系统基础
Java / Spring Boot / TypeScript / Angular / React / SQL / CI/CD
云与数据执行
AWS / S3 / Lambda / ECS / CloudWatch / RDS / DynamoDB / Athena-style analytics
Applied AI 落地
LLM workflows / evals / quality gates / observability / approval systems
产品判断
Private deployment / data safety / operator workflows / measurable demos
- 把 AI 系统做进真实业务工作流
- 把 LLM Demo 变成可评测、可观测、有人类审批的应用
- 为重视数据安全的业务方设计本地/私有 AI 工具
- 用开源项目和内容持续展示实际 AI 工程能力
English
Not another chatbot wrapper. I build the control plane around AI: data flow, evaluation, approval gates, observability, deployment, and the product surfaces people actually use.
I am a senior full-stack/backend engineer moving deeply into applied AI delivery. My work focuses on the missing layer between AI demos and production systems: evaluation, observability, workflow design, approval gates, AWS-backed data flows, and deployable product surfaces.
| What I bring | Why it matters |
|---|---|
| 10+ years in enterprise software | I understand real systems, release pressure, data boundaries, production support, and team coordination. |
| Java, TypeScript, AWS, data workflows | I can connect AI features to existing enterprise stacks instead of leaving them as isolated demos. |
| LLM evals, observability, approval gates | I design AI systems teams can test, monitor, and govern. |
| Product-minded side projects | I build toward usable workflows, not just experiments. |
| System | What it proves | Link |
|---|---|---|
| JTestGen AI Java test generation |
Developer productivity system with JaCoCo targeting, JUnit/Mockito generation, Maven validation, repair loops, and CI-ready artifacts. | Repo |
| DREAM Source-backed team memory platform |
Open-core / private-extension architecture for source-backed AI engineering workflows: memory packs, codebase indexes, Evidence Graph, governed memory distillation, Requirement Cases, PR review summaries, scorecards, provenance/security boundaries, and SQLite audit trails. | Repo |
| Enterprise AI Engineering Orchestrator Release-risk and approval workflow |
Shows auditable engineering AI orchestration with repository context, risk scoring, test recommendations, UiPath Test Manager proof, DynamoDB proof, and human approval gates. | Repo · Demo · Proof |
| LLM Eval Observability LLM evaluation and observability |
Repeatable AI app checks with provider adapters, quality gates, SQLite run history, and evaluation-first engineering. | Repo |
| AI ContentOps Studio AI content workflow |
Traceable content operations: research, generation, quality scoring, human review, release evidence, and deployment config. | Repo |
| AI Commerce Copilot Cross-border commerce AI assistant |
Business AI workflow design: product scoring, supplier quote loops, content drafts, approval gates, audit evidence, and private/local deployment boundaries. | Portfolio |
| Brand Publishing Pipeline Technical brand publishing workflow |
Technical content ops for blogs, short-video scripts, platform metadata, review gates, and safe publishing handoff. | Repo |
Enterprise systems foundation
Java / Spring Boot / TypeScript / Angular / React / SQL / CI/CD
Cloud and data execution
AWS / S3 / Lambda / ECS / CloudWatch / RDS / DynamoDB / Athena-style analytics
Applied AI delivery
LLM workflows / evals / quality gates / observability / approval systems
Product judgment
Private deployment / data safety / operator workflows / measurable demos
- Building AI systems that survive contact with real business workflows
- Turning LLM demos into evaluated, observable, human-approved applications
- Designing private/local AI tools for operators who care about data safety
- Using content and open-source projects to show practical AI engineering depth
- Portfolio: zemeng2015.github.io/zack-ai-homepage
- LinkedIn: linkedin.com/in/zack-wang-profile
- Email: zackwang.job@gmail.com