AI Architecture & Automation | Supply Chain Specialist AI 架构与自动化 | 供应链专家

Hi, I'm Tong Wu.
I build AI agent systems
and intelligent supply chains.

你好，我是吴桐。
我构建 AI 智能体系统，
用 AI 改变供应链效率。

M.Sc. Economics & Finance · Six Sigma Green Belt · AI Agent Architect. I ship production MCP servers, multi-agent orchestration pipelines, and agent memory systems — while bringing deep supply chain expertise from Amazon, Xiaomi, and Huawei.

经济与金融学硕士 · 六西格玛绿带 · AI 智能体架构师。我发布生产级 MCP 服务器、多智能体编排流水线和智能体记忆系统，同时拥有来自亚马逊、小米和华为的深厚供应链实战经验。

Explore My Work查看我的作品 Contact Me联系我

About Me

关于我

I specialize in bridging supply chain operational expertise with AI agent architecture. Currently building production AI tools at Novada Labs — MCP servers, multi-agent pipelines, and agent memory systems published on npm and used by developers worldwide.

Previously optimized omni-channel B2C/B2B fulfillment for global tech leaders: Amazon, Xiaomi, and Huawei. At Amazon I leveraged company Claude AI for operational automation; at Xiaomi I integrated ChatGPT into S&OP workflows.

I engineer customized AI agents using Python, Claude Code, Codex, OpenClaw, and MCP. With deep understanding of agent architecture — planning, memory, tool-use, multi-agent coordination — I build systems that work autonomously overnight and deliver measurable output.

This dual background lets me identify strategic pain points and personally implement the technical solutions, from supply chain SOP to agent orchestration framework.

我致力于将供应链运营专长与 AI 智能体架构能力深度融合。目前在 Novada Labs 构建生产级 AI 工具：MCP 服务器、多智能体流水线和智能体记忆系统，已发布到 npm 并被全球开发者使用。

此前为亚马逊、小米和华为等全球科技巨头优化全渠道 B2C/B2B 订单履行。在亚马逊使用公司 Claude AI 实现运营自动化；在小米将 ChatGPT 整合至 S&OP 工作流。

使用 Python、Claude Code、Codex、OpenClaw 和 MCP 构建定制化 AI 智能体。深入理解智能体架构（规划、记忆、工具调用、多智能体协调），构建可过夜自主运行并产出可量化结果的系统。

这种双重背景使我不仅能精准洞察业务痛点，还能亲手实现底层技术方案——从供应链 SOP 到智能体编排框架。

Location所在地
Düsseldorf, Germany (EU Blue Card)
Education教育背景
M.Sc. Economics & Finance
(University of Freiburg)
Languages语言能力
Chinese (Native), English (C2), German (C1), Japanese (N4) 中文(母语), 英语(C2), 德语(C1), 日语(N4)
Certifications专业认证
Six Sigma Green Belt · EU Driver's License 六西格玛绿带 · 欧盟驾照

Work Experience

工作经验

Mar 2026 – Present

AI Automation & Architecture Engineer

AI 自动化与架构工程师

Novada Labs — Remote

  • MCP Product Engineering: Architected and shipped novada-search, a production MCP server with 11 tools (search, extract, crawl, map, scrape, browser) published on npm (v0.8.3). Built novada-proxy-mcp (8 tools, 241 automated tests) scoring 8.8/10 on agent-first evaluation vs. BrightData (4.5), Oxylabs, and Decodo — highest in category.
  • Multi-Agent Orchestration (AAM): Designed Agent Autonomous Mode — a 4-stage production pipeline using Claude Opus as orchestrator dispatching parallel Sonnet workers, with independent Reviewer and Verifier agents. Enables overnight autonomous task execution with FSM state management and budget controls. Claude Code Codex
  • Agent Memory System (AgentRecall): Built and published AgentRecall (npm v3.3.28) — an MCP-native agent memory layer with correction-based calibration, cross-session recall, and long-term knowledge consolidation. Adopted as core infrastructure across all production agent workflows.
  • Benchmark Testing: Designed a 200-URL automated benchmark framework validating novada-search vs. Firecrawl, Tavily, and BrightData across 6 purchase-decision dimensions. Identified Firecrawl's 33% anti-bot failure rate as key competitive differentiator.
  • Ecosystem Distribution: Shipped CLI, Python/npm SDKs, and Claude Code skills for both products. Submitted Claude Plugin Marketplace listing. Completed LangChain and CrewAI integrations; built n8n workflow nodes and automated competitive intelligence pipeline using scheduled agents.
  • MCP 产品工程：架构并发布 novada-search，一个拥有 11 个工具（搜索、提取、爬取、映射、抓取、浏览器）的生产级 MCP 服务器（npm v0.8.3）。构建 novada-proxy-mcp（8 个工具，241 个自动化测试），智能体优先评分 8.8/10，高于 BrightData（4.5）、Oxylabs 和 Decodo。
  • 多智能体编排（AAM）：设计智能体自主模式——4 阶段生产流水线，使用 Claude Opus 作为编排器调度并行 Sonnet 工作智能体，配备独立的审查和验证智能体。支持过夜自主任务执行，含 FSM 状态管理和预算控制。Claude Code Codex
  • 智能体记忆系统（AgentRecall）：构建并发布 AgentRecall（npm v3.3.28）——基于 MCP 的智能体记忆层，含纠错校准、跨会话召回和长期知识宫殿整合。已作为所有生产智能体工作流的核心基础设施。
  • 基准测试：设计 200 URL 自动化基准框架，在 6 个购买决策维度上对比 novada-search 与 Firecrawl、Tavily 和 BrightData。识别出 Firecrawl 33% 的反爬失败率作为关键竞争护城河。
  • 生态分发：为两个产品发布 CLI、Python/npm SDK 和 Claude Code 技能。提交 Claude 插件商城上架申请。完成 LangChain 和 CrewAI 集成；构建 n8n 工作流节点和自动化竞争情报流水线。
Sep 2025 – Jan 2026

Process Excellence Manager

流程精益经理

Amazon — Hamburg-Nutzen (DHH2)

  • Last-Mile Order Fulfillment & Leadership: Co-managed daily station operations for B2C last-mile fulfillment, directing 100+ L1-L3 associates. Conducted RCA and executed monthly process, safety, and compliance audits, maintaining strict 100% safety compliance.
  • DSP Management & Process Excellence: Evaluated 16 DSPs via data-driven KPI analysis (100+ OTR/UTR metrics). Sustained a 99.5% DEA rate and translated KPI insights into Lean process improvements, boosting station labor productivity.
  • AI-Augmented Workflow Automation: Built and deployed custom automation tools (Tampermonkey + VBA + company Claude AI) to streamline data extraction, RCA documentation, and operational reporting. Leveraged Claude for automated KPI narrative generation and incident summarization — saving the management team 6+ hours/week. Claude AI
  • 最后一公里订单履行与团队领导：共同管理配送站日常运营，指导 100+ 名 L1-L3 员工。执行 RCA 及月度流程、安全与合规审计，维持 100% 安全合规。
  • DSP 管理与流程卓越：通过数据驱动方式评估 16 家 DSP（100+ OTR/UTR KPI），维持 99.5% DEA 率，将 KPI 洞察转化为精益改进，显著提升站点劳动力生产率。
  • AI 增强工作流自动化：部署自定义自动化工具（Tampermonkey + VBA + 公司 Claude AI），简化数据提取、RCA 文档和运营报告。利用 Claude 自动生成 KPI 叙述和事件摘要，每周为管理团队节省 6+ 小时。Claude AI
Jun 2024 – Aug 2025

Supply Chain Delivery Manager

供应链交付经理

Xiaomi Technology — Düsseldorf

  • End-to-End Order Fulfillment & Logistics: Managed B2B/B2C omnichannel fulfillment (≈320k units/year, ~€110M) for Key Accounts in Germany and the Netherlands. Directed international air freight, customs clearance, and local 3PL distribution via CRM, GDP, and SAP, achieving a 99.5% OTD rate.
  • AI-Assisted Demand Planning & S&OP: Led CPFR demand planning and SKU lifecycle management using Excel analytics augmented by ChatGPT for structured demand narrative generation and S&OP documentation. Translated insights into 12-month rolling supply plans; optimized ITO and aligned GTM/Sales against inflated KA forecasts. ChatGPT
  • Key Account Strategy & Profitability Analysis: Evaluated holistic KA profitability (unit sales, product mix, logistics costs). Formulated annual KA strategies and executed complex product rework projects (custom software/packaging) for Deutsche Telekom and Vodafone.
  • Proprietary Delivery Platform: Designed and built an internal delivery tracking platform adopted company-wide, streamlining KPI visibility and exception management (damages, delays, mis-shipments). Established SOPs reducing operational risks and logistics costs continuously.
  • 端到端订单履行与物流：管理德国和荷兰大客户 B2B/B2C 全渠道履行（约 320k 台/年，约 1.1 亿欧元）。通过 CRM、GDP 和 SAP 管理国际空运、清关和本地 3PL，实现 99.5% OTD 率。
  • AI 辅助需求规划与 S&OP：使用 Excel 数据分析结合 ChatGPT 主导 CPFR 需求规划和 SKU 生命周期管理，自动生成结构化需求叙述和 S&OP 文档。将洞察转化为 12 个月滚动供应计划，优化 ITO。ChatGPT
  • 大客户战略与盈利分析：评估 KA 整体盈利能力（单位销量、产品组合利润率、物流成本）。制定年度 KA 战略，并为德国电信和沃达丰执行复杂产品返工项目。
  • 专属交付平台：设计并构建内部交付追踪平台，被整个供应链团队采用，简化 KPI 可见性和异常管理（损坏、延误、错发）。建立持续降低运营风险和物流成本的 SOP。
Dec 2023 – Jun 2024

EU Certification Specialist

欧洲认证专员

Huawei — Düsseldorf

  • Rebuilt the Western Europe certification platform; managed Q-EHS (ISO 9001/14001/45001) and ESG compliance modules.
  • 重建西欧区认证平台；管理 Q-EHS（ISO 9001/14001/45001）及 ESG 合规模块。
Jul 2021 – Oct 2021

Business Intern

商务实习生

SEG Automotive — China

  • Executed procurement processes within SAP and assisted in supplier risk management and safety audits.
  • 在 SAP 系统中执行采购流程，协助供应商风险管理和安全审计。
Nov 2018 – Mar 2020

Foreign Trade Specialist

外贸专员

Arshine Pharmaceutical Co., Ltd — Changsha

  • Expanded overseas client base and coordinated third-party international logistics for B2B customers.
  • 开拓海外客户，协调第三方国际物流，为 B2B 客户提供产品供应。
Jun 2018 – Sep 2018

Business Intern

商务实习生

Harris Teeter — USA

  • Supported retail operations in logistics and sales in a high-volume retail environment.
  • 在高业务量零售环境中支持物流和销售运营工作。

University of Freiburg

弗莱堡大学

Teaching & Research Assistant (HiWi) · Oct 2022 – Feb 2024 · Freiburg, Germany 助教与研究助理 · 2022年10月 – 2024年2月 · 德国弗莱堡

Tutor — Advanced Macroeconomics I & II · Econometrics · Principles of Finance 助教——高级宏观经济学 I & II · 计量经济学 · 金融学原理
Python, Z-tree, LaTeX, Data Analysis
Research Assistant — UBI & Game Theory 研究助理——全民基本收入与博弈论
R, Z-tree

AI Agent & Product Innovation

AI 智能体与产品创新

May 2025 – Jun 2025

Product Designer & Developer — Xiaomi Hackathon

产品设计与开发 — 小米黑客松

AI Multi-Modal Smart Thermometer

AI 多模态智能温度计

  • Integration: Led concept design for an AI-powered parenting thermometer integrating Whisper (infant cry interpretation) and GPT-4 Vision (stool image analysis); independently built interactive frontend pitch demo.
  • Architecture: Designed end-to-end hardware-to-cloud pipeline: ESP32 sensors → BLE 5.0 → Cloud LLM processing, translating the AI pipeline into a coherent user story and business case.
  • 创新整合：主导 AI 育儿温度计概念设计，整合 Whisper（婴儿哭声识别）与 GPT-4 Vision（粪便图像分析），独立构建交互式前端演示。
  • 技术架构：设计全链路架构：ESP32 传感器 → BLE 5.0 → 云端大模型处理，将 AI 能力转化为完整商业方案。
2024 – Present

Independent Developer — Multi-Agent Systems & Open Source

独立开发者 — 多智能体系统与开源

Agent Architecture, Orchestration & Tooling

智能体架构、编排与工具链

  • 4-Stage Pipeline: Engineered production multi-agent pipelines using Claude Code, Codex, OpenClaw, and MCP — Orchestrator (Claude Opus) dispatches parallel Worker agents (Sonnet), with independent Reviewer and Verifier agents catching errors workers miss. This pattern consistently outperforms single-agent execution on complex, multi-day tasks.
  • Ecosystem Integration: Applied deep understanding of agent architecture (planning, memory, tool-use, function calling) to build scalable workflows deployable on Coze, Dify, n8n, LangChain, and CrewAI. Built automated knowledge retrieval, cross-system orchestration, and MCP server tooling from scratch.
  • 4 阶段流水线：使用 Claude Code、Codex、OpenClaw 和 MCP 构建生产级多智能体流水线——编排器（Claude Opus）调度并行工作智能体（Sonnet），配备独立审查器和验证器捕获工作智能体遗漏的错误。该模式在复杂多日任务上持续优于单智能体执行。
  • 生态集成：深入理解智能体架构（规划、记忆、工具调用、函数调用），构建可部署在 Coze、Dify、n8n、LangChain 和 CrewAI 上的可扩展工作流。从零构建自动化知识检索、跨系统编排和 MCP 服务器工具链。

Portfolio & Documents

作品库与文档阅览

Academic papers, custom code scripts, certifications, and professional references. PDFs open in a new tab. Code opens in an interactive viewer.

学术论文、代码脚本、专业资质证书和推荐信。点击 PDF 在新标签页预览，点击代码在弹窗中高亮显示。

Programming & Code

编程与代码项目

Python & JavaScript Automation
Python 与 JS 自动化脚本

Automation scripts integrating AI models — Tampermonkey for Amazon internal systems and a LangChain RAG PDF reader.

整合 AI 大模型的自动化脚本，包含亚马逊内部系统 Tampermonkey 增强脚本和 LangChain RAG PDF 阅读器。

TM Script GitHub Pinecone.py PDF App.py

Academic Research

学术研究论文

University of Freiburg
弗莱堡大学

Economics × technology intersections: UBI in the AI era and socio-economic impacts of the GPT ecosystem.

经济学与前沿科技交叉研究：AI 时代的全民基本收入（UBI）以及 GPT 生态系统的社会经济影响。

Master's Thesis GPT Ecosystem

Certifications

专业资质认证

Coursera & Institutions
Coursera 及专业机构

Six Sigma, Data Analytics, Python Automation, Machine Learning, and Business Analysis credentials.

六西格玛绿带、谷歌数据分析、Python 自动化、斯坦福机器学习及高级商业分析（CBAP）资质证明。

View All

Reference Letters

工作与学术推荐信

Professional Endorsements
来自雇主与教授的认可

Official letters from Xiaomi, Huawei (Adecco), AEEC, and professors at the University of Freiburg.

来自小米、华为、AEEC 的官方推荐信，以及弗莱堡大学教授的学术推荐信。

View Letters

Core Competencies

核心技能矩阵

AI Architecture & Agent Engineering

AI 架构与智能体工程

AI Agent Building | 智能体开发 MCP / Claude Code / OpenClaw Codex | 代码生成 Multi-Agent Orchestration | 多智能体编排 Python (Workflow Auto) | 自动化工作流 Vector DB / RAG | 向量检索增强 Prompt Engineering | 提示词工程 LangChain / CrewAI n8n / Coze / Dify VBA / Tampermonkey LLM: GPT / Claude / Gemini / Deepseek

Strategic Operations & Supply Chain

战略运营与供应链管理

Last-Mile Logistics | 最后一公里物流 S&OP / Demand Planning | 产销协调与需求规划 DSP Management | 配送服务商管理 Six Sigma Green Belt | 六西格玛绿带 Omnichannel Retail Ops | 全渠道零售运营 SAP / ERP Systems

Data Intelligence & Analytics

数据智能与分析

Business Intelligence | 商业智能分析 Econometrics & ML | 计量经济与机器学习 R / Python / Excel Advanced

Interests & Philosophy

个人爱好与哲学探索

Beyond algorithms and supply chains, I have a deep fascination with traditional Eastern metaphysics. Ancient wisdom on change and balance offers a unique perspective for navigating modern complexity.

在算法与供应链的严谨世界之外，我对东方传统玄学有深厚兴趣。古代关于"变易"与"平衡"的智慧，为现代复杂系统提供了独特的降维视角。

BaZi (Four Pillars)

八字命理

Studying the complex interactions of the Five Elements (Wu Xing) to understand natural life cycles and systemic balance.

研究五行生克制化的复杂互动，探索生命周期与自然规律的底层逻辑与平衡之道。

I Ching (Book of Changes)

易经研读

Exploring ancient hexagrams to grasp the philosophy of constant change, adapting strategic thinking in dynamic environments.

通过卦象推演理解"变易"哲学，将古人战略决策思维应用于现代动态环境分析。

Divination Arts

术数与占卜

Using traditional divination as a psychological and reflective tool to approach problems from non-linear, intuitive angles.

将传统术数作为心理投射与自我反思的工具，跳出线性逻辑，从直觉和全息视角拆解难题。