Program manager with a builder's instinct. At Foxconn, I owned FATP (final assembly) material readiness for Apple's iPhone program — a single build spanned 12 lines, roughly 1,500 unique part numbers and 45+ suppliers. Beyond running the process, I used AI to design and build the tools it needed. Before that, I coordinated NTD-billion-scale construction projects, where I learned the same instinct: catch problems before they cost time and money.
有工程師底子的專案經理:曾在鴻海負責 Apple iPhone 專案 FATP 最終組裝的備料,單一建置橫跨 12 條產線、約 1,500 個不重複料號、協調 45+ 家供應商。除了執行流程本身,也利用 AI 設計並開發流程需要的工具。建築本科與 56 億營建專案的經驗,讓我具備掌握大型專案的全局觀,習慣在問題發生前就先擋下來。
Understand how a part's data moves from the dock through the supplier system, SAP, and into iMES — the basis for eliminating input errors at the root.理解一顆料從碼頭進來、經供應商系統、SAP,再到 iMES 的完整路徑,從根本排除工廠投入錯誤。
Cross-functional translation跨部門翻譯與橋接
Coordinate engineering, manufacturing, and supply chain across every NPI build cycle — teams that don't naturally speak the same language.在每個 NPI 建置週期協調工程、產線、供應鏈,在不容易對話的團隊之間做翻譯。
Architecture-rooted rigor建築背景帶來的嚴謹
Original background is architecture — reviewing structural, MEP, and site-condition drawings to resolve multi-trade conflicts before they reach the field.本科是建築,審查結構、機電圖說與現場條件,在問題進場前跨工種化解衝突。
Builds the tools, not just the process不只跑流程,也做工具
Designs and builds the automation the workflow needs.設計並開發流程需要的自動化工具。
Time-lapse · shot and edited by Deson Quach縮時攝影架設及編輯 by 郭泓
Three problems, three different kinds of leverage三個問題,三種不同的解法邏輯
Dev Tools開發工具
A production-scheduling automation tool, built with AI跟 AI 協作做出來的生產排程自動化工具
Built and maintain a single-file, zero-backend browser tool (~11,600 lines of hand-evolved HTML/JavaScript) that drives daily production-scheduling and parts-management workflows. Developed iteratively through natural-language collaboration with an AI coding assistant — acting as requirements owner, domain expert, and quality gate, not the one writing every line.
獨立設計並持續維護一套單檔案、零後端的網頁自動化工具(約 1.1 萬行 HTML/JavaScript),處理每天的生產排程與料件管理工作。透過跟 AI 程式助理的自然語言對話持續迭代開發——角色是提出需求、驗證邏輯正確性,不是自己寫每一行程式。
Recommendation engine推薦引擎Custom part-similarity scoring + live inventory + dependency rules to rank replacement parts.自己設計的字串相似度演算法+即時庫存+跨部門相依性規則,排出最佳替代料件順序。
Diff engine差異偵測引擎Catches silent changes between report versions — 300+ real discrepancies per run.自動比對新舊報表,單次可抓出 300+ 筆被悄悄修改的差異。
Dependency checks相依性檢查Cross-validates naming and selection rules a person would likely miss.自動核對命名與選料規則,抓出人工容易漏看的邏輯衝突。
Debugging discipline除錯紀律Found & fixed a dozen+ "silent legacy-assumption" bugs through systematic live-data verification.靠系統性拿真實資料驗證,抓出並修復十幾個「舊假設靜默失效」的邏輯錯誤。
If asked which client/product this is for: under NDA, I can't share specifics, but I'm glad to walk through the technical approach. All screenshots below use fake sample data.被問「這是幫哪家客戶做的」:因為保密協議不能透露客戶細節,但可以講技術做法。以下所有截圖均為假資料重新操作拍攝。
Drop a schedule export in, and the tool parses it instantly.拖一份排程報表進去,立即解析出可操作的介面。Special List — auto-generated exception parts list.Special List——自動產出的特殊料件清單。Board-matching — daily mainboard-variant demand.板端對料——當天主板種類與數量需求。Next-day input report with an auto-written summary.明天投入日報,附自動生成的文字摘要。Dependency check — works alongside customer verification testing to comprehensively confirm correct part selection.相依性檢查——配合客戶驗證測試,全方位確認是否有正確選料。Repair/replacement recommendation engine — full walkthrough.維修換料推薦引擎——完整操作流程。
