ABOUT
專門處理已在 production 運作、不能輕易重來的複雜系統。Systems engineering for complex production systems that cannot simply start over.
我主要處理 legacy system、高併發、production incident、效能瓶頸與高風險系統變更。這些問題橫跨應用程式、資料庫、網路與基礎設施，需要在既有系統的限制裡，找出安全、可驗證且能實際執行的下一步。I focus on legacy systems, high concurrency, production incidents, performance bottlenecks, and high-risk changes. These problems span applications, databases, networks, and infrastructure. My work is to find a safe, verifiable, practical next step within the constraints of the existing system.
我會先把已知事實、假設與缺少的觀測分開，找出下一個能縮小範圍的實驗。從診斷、設計到實作，保留每次判斷的依據，讓團隊能自己覆核，也能接著往下做。I separate known facts, hypotheses and missing observations, then identify the next experiment that can narrow the scope. From diagnosis through design and implementation, I preserve the reasoning so the team can review it and continue the work.
工作範圍SCOPE OF WORK
主要處理focusLegacy / 高併發 / Productionlegacy / concurrency / production
合作型態engagement診斷 / 專案 / 長期顧問diagnosis / project / retained advisory
語言languages中文 / English
保密confidentiality客戶系統與合作內容不對外公開；公開案例僅使用去識別化、重新整理後的示範情境。Client systems and engagements remain confidential. Public examples are anonymized, reworked illustrative scenarios.
JUDGMENT PRINCIPLES
我下判斷時依據的原則。The principles behind how I decide.
先量測，再改動。Measure before changing anything.
先建立可比較的基準線；必要時補最關鍵的觀測資料，避免在證據不足時直接改動系統。Establish a comparable baseline and fill critical observation gaps before changing a system on insufficient evidence.
最小可回滾的改動優先。The smallest reversible change comes first.
在 production 上，能安全退回來的方案，比理論上更完整的方案有用。In production, a change you can safely back out of is worth more than a theoretically cleaner one.
保留判斷，不只交付結果。Hand over the reasoning, not just the result.
決策理由都會寫下來，讓團隊之後能自己覆核、自己延伸，不必回頭問我。Every decision is written down so the team can review it and build on it later without coming back to me.
尊重既有系統的理由。Respect the reasons the system is the way it is.
看起來不合理的設計，背後可能存在當時真實的限制；改動前先理解它為什麼存在。An unusual design may reflect a real constraint at the time. Understand why it exists before changing it.
說得出代價，才算給了建議。Advice without a stated cost is not advice.
只講好處的不算建議。每個選項都會附上它讓什麼變差。Every option comes with what it makes worse, not only what it improves.
不適合的案子直接說。Say so when the work is not a fit.
問題不在我的能力範圍，或其實不需要我，我會直接說，並給一個方向。If the problem is outside what I do, or you do not actually need me, I will say it and point somewhere useful.
適合找我的情況A GOOD FIT
系統會慢、會出錯，但沒有人指得出原因。The system is slow or fails, and nobody can point at the cause.
有人提議重寫，你不確定值不值得。A rewrite has been proposed and you are not sure it is worth it.
系統必須持續演進，同時不能停機。The system has to keep evolving and cannot go down.
同樣的事故一直重複，每次都靠人的經驗擋過去。The same incident keeps recurring and is held off each time by someone's experience.
要做資料搬遷或架構調整，但擔心風險收不住。A migration or architectural change is needed and the risk feels uncontained.
不適合的情況NOT A FIT
需要大量人力同時開發多個新功能。You need headcount to build several new features in parallel.
需要純前端或視覺設計的產出。You need front-end or visual design work.
只以最低價格完成指定功能為目標。The sole goal is to implement specified features at the lowest price.
只需要按固定規格補充開發人力，不涉及系統判斷。You only need extra development capacity against a fixed specification, without systems judgment.
完全無法提供去識別化的問題描述、觀測資料或必要背景。No anonymized problem description, observations or necessary context can be shared.