AARON LIANG / A PORTRAIT

人物誌・2026PERSONAL PROFILE · 2026

把複雜,
做成日常。
Making complex technology useful in everyday life.

從晶片裡的微小元件,到金融產品,再到一間門市的日常決定。領域一直在變,我在意的事始終一樣:讓技術真的幫上忙。From the smallest cells in a chip to financial products and shop-floor decisions, the fields keep changing. What stays the same is what matters to me: making technology truly useful.

梁家為Aaron Liang

電機與電腦工程博士/產品經理/AI 實踐者Ph.D. in Electrical and Computer Engineering · Product Manager · AI Practitioner

梁家為 Aaron 穿著米色西裝的半身形象照
梁家為 · Chia-Wei (Aaron) LiangChia-Wei (Aaron) Liangfig. 01

Always curious. Always building.

翻開我的故事Read my story

第一章・晶片Chapter One · Chips

故事,從最小的地方開始。A story that starts small.
Very small.

在晶片的世界裡,一條設計規則、一個元件的佈局,都必須經得起驗證。這也是我思考問題的起點。In chip design, every rule and every cell layout must withstand verification. That is where my way of thinking began.

我在交通大學讀電機與電腦工程,之後繼續攻讀博士。研究的問題很具體:能不能把專家才懂的設計規則,變成電腦能理解、能自動執行的方法?I studied Electrical and Computer Engineering at NCTU, then pursued a Ph.D. at NYCU. My question was concrete: could the design rules that only experts understand become something a computer can understand and apply?

2019 年到聯發科做 EDA,讓這個問題走出了實驗室。我開發的元件 layout 方法與流程,後來導入了實際產線,用在出貨的產品線上。Joining MediaTek as an EDA Engineer in 2019 took that question beyond the lab. The cell-layout methods and flow I developed were adopted in MediaTek’s production flow to generate cells used in shipping products.

博士論文 AutoCell,將商用 5 奈米元件庫中的 900 多個組合邏輯元件,在約 7 分鐘內全數生成,沒有 DRC/LVS 違規。對我來說,研究最有意義的時刻,是它開始真正運作的時候。My dissertation, AutoCell, generated 900+ combinational cells from a commercial 5nm library in about seven minutes, with zero DRC/LVS violations. For me, research becomes most meaningful when it starts working in the real world.

2024 年起,我在陽明交大擔任博士後研究員,持續參與聯發科產學 AI 研究,也指導研究生。研究與實務,始終是同一條路。Since 2024, I have been a postdoctoral researcher at NYCU, continuing industry–academia AI research with MediaTek and mentoring graduate students. For me, research and practice have always been the same road.

讀研究與論文Explore the research
梁家為獲頒 IEEE 台北分會最佳博士論文獎
01 / IEEE 台北分會 2024 最佳博士論文獎01 / IEEE Taipei Section 2024 Best Ph.D. Dissertation Award
博士論文口試合照
02 / 博士口試,一段研究旅程的里程碑。02 / The dissertation defense: a milestone in my research journey.
14

篇同儕審查論文(2019–2026)
IEEE TVLSI、TCAD、DAC、ICCAD、DATE 等;包含第一作者及共同作者研究。
Peer-reviewed papers, 2019–2026
IEEE TVLSI, TCAD, DAC, ICCAD, DATE, and others; includes first-author and co-authored research.

2012–2016國立交通大學・電機與電腦工程學士NCTU · BS, Electrical and Computer Engineering從工程基礎出發。A foundation in engineering.

2016–2024國立陽明交通大學・電機與電腦工程博士NYCU · Ph.D., Electrical and Computer Engineering在研究與產業之間,反覆驗證想法。Testing ideas in both research and industry.

第二章・金融科技Chapter Two · Fintech

從做好一個元件,
到帶著一群人做產品。
From building a cell
to building with a team.

2022 年,我成為 TKSpring(Term Structure)的第 2 號員工。從工程師走到產品經理,問題的尺度變了,對細節的要求沒有。In 2022, I became Employee #2 at TKSpring (Term Structure). As I moved from engineer to Product Manager, the problems grew in scale, but my attention to detail remained the same.

TKSpring 團隊在辦公室拿著春聯合照
03 / TKSpring・一起把產品做出來的人。03 / TKSpring · The people behind the products.People make it possible.

團隊從 5 人擴編到 30 人,後來整併為 15 人。我負責三代固定利率借貸產品的設計、開發與上線,也學會在技術選擇、產品方向與團隊協作之間做判斷。The team grew from 5 to 30 people, later consolidating to 15. I led the design, development, and launch of three generations of fixed-rate lending products, making decisions across technology, product direction and teamwork.

TermMax 橫跨 6 條鏈,歷史鎖倉規模(TVL)峰值超過 1 億美元。這是整個團隊的產品成果。而我的責任,也包含那些看起來很小、卻不能出錯的事。TermMax reached a historical peak TVL above $100 million across six chains—a product outcome achieved by the team. My responsibilities also included the small details that could not be allowed to go wrong.

TMX 上線前,我把 58,974 條代幣釋放流逐筆核對,抓出了 3 個錯誤。從晶片到金融,驗證這件事,從來沒有離開過我的工作。Before the TMX launch, I reconciled all 58,974 vesting streams one by one and caught three errors. From chips to finance, verification has stayed at the heart of my work.

看三代產品的完整故事Read the product story
  1. I

    Term Structure

    zk rollup 訂單簿。當架構限制了可組合性,我主導下線與轉型的決定。A zk-rollup order book. When its architecture limited composability, I led the decision to sunset it and pivot.

  2. II

    TermMax

    全鏈上固定利率借貸,2025 年上線,拓展至 6 條區塊鏈。Fully on-chain fixed-rate lending, launched in 2025 and expanded across six blockchains.

  3. III

    TermPrime

    Canton Network 上的機構級附買回,2026 年 6 月完成首筆正式交易。Institutional repo on Canton Network, with its first production trade completed in June 2026.

把複雜的技術,做成真的有人用、
真的有成效的東西。
Turn complex technology into products
people actually use—with measurable results.

梁家為/我在不同領域裡,一直在做的事AARON LIANG / THE COMMON THREAD IN MY WORK

第三章・走進門市Chapter Three · On the Shop Floor

先聽懂日常,
再談數位轉型。
Understand daily work first.
Then discuss digital transformation.

2026 年,我開始擔任零售數位轉型顧問。到了門市,問題變得很生活:什麼賣得好?這檔活動真的有效嗎?老闆明天該怎麼決定?In 2026, I began advising retail businesses on digital transformation. On the shop floor, the questions are everyday ones: What sells well? Did this campaign work? What should the owners do tomorrow?

我的做法是先到現場,看老闆每天怎麼做決定,再用最輕的工具把資料接起來。客戶不用學新軟體,打開手機就看得到。I start by watching how owners make daily decisions, then connect the data using simple, lightweight tools. They can see what they need on their phones, without learning new software.

逐檔分析 47 檔行銷活動,POS 資料連續自動同步 378 天、零缺口。合作也從單次專案,走成了年度顧問約。I analyzed 47 campaigns individually and kept POS data syncing for 378 consecutive days without gaps. A one-off project became an annual advisory engagement.

看 AI 如何走進工作現場See AI at work

門市的一天,資料怎麼走到店長手機A day of shop data, from till to phone

示意數據Sample data
  • 30 秒 secPOS 數字自動同步到手機POS numbers sync to the phone automatically
  • 3 邊 sources記帳表、POS、銀行每天自動對帳,對不上的日子標紅Ledger, POS and bank reconciled daily; mismatched days flagged red
  • 44 份 reports每週自動週報與檔期復盤;30 個月 POS 資料、6 類手法算勝率Weekly reports and campaign reviews from 30 months of POS data, scoring 6 types of tactics

示意圖與示意數據,不代表任何特定門市。Illustration with sample data; not tied to any specific shop.

第四章・第二大腦Chapter Four · A Second Brain

要教別人用 AI,
就先把它放進自己的生活。
Before teaching others to use AI,
I make it part of my own life.

魔法小島Magic Island
魔法小島:把第二大腦變成每天想打開的地方(畫面為示範資料)Magic Island: turning a second brain into a place you want to open every day (demo data)
黃昏Dusk
夜晚・極光Night · aurora

同一座島,
隨時間改變樣子。
The same island,
changing with the hours.

我的第二大腦,是自己建置、每天都在用的個人系統。會議錄音、郵件、行事曆與待辦,自動整理成可以搜尋的知識,也變成能直接採取行動的清單。My second brain is a personal system I built and use every day. Meeting recordings, emails, calendars, and to-dos become searchable knowledge and lists I can act on.

我可以在手機上直接問、直接派工。背後有長期記憶、工具調度、權限管理與故障恢復。對外建立東西之前,系統會等我按下核准;方便之外,也要知道界線在哪裡。I can ask questions and delegate from my phone. Behind it are long-term memory, tool orchestration, permissions, and recovery. Before creating anything in an external system, it waits for my approval. Convenience needs boundaries.

打開我的第二大腦Inside my second brain
  1. 01 /

    留下來Capture

    錄音轉文字,郵件與行程同步。Transcribe recordings; sync emails and calendars.

  2. 02 /

    連起來Connect

    會議、決定與人物,歸檔成知識。Organize meeting records, decisions, and information about people into a knowledge base.

  3. 03 /

    問得到Retrieve

    透過 AI 助理,隨時找回脈絡。Retrieve context through an AI assistant, anytime.

  4. 04 /

    做得到Act

    每日簡報、待辦追蹤與逾期提醒。Daily briefings, task tracking, and overdue reminders.

第五章・講台上Chapter Five · On Stage

挖得更深一點,
再講給大家聽。
Keep digging deeper.
Then share what I find.

我喜歡持續研究更深、更新的技術:從晶片設計自動化、區塊鏈的零知識證明,到現在的 AI agent。每弄懂一層,就想把它講給大家聽,不管是在 DAC、ICCAD 這類國際學術研討會、大學課堂,還是企業 AI 工作坊。I love digging into deeper, newer technology: from chip design automation to zero-knowledge proofs in blockchain, and now AI agents. Each time I understand another layer, I want to share it, whether at international academic conferences such as DAC and ICCAD, in university classrooms, or in company AI workshops.

11

篇論文,發表在國際學術研討會
另有 3 篇 IEEE 期刊論文(TCAD、TVLSI)
papers at international academic conferences
plus 3 IEEE journal papers (TCAD, TVLSI)

  1. ITC
  2. DATE
  3. VLSI-DATITC-AsiaITC
  4. ITC
  5. ICCAD
  6. ICCADITC
  7. ASP-DACDAC

★ 親自上台報告:ITC 2019、DATE 2021(第一作者)、DAC 2026(美國長灘)★ Presented in person: ITC 2019, DATE 2021 (first author), DAC 2026 (Long Beach)

研究做到一個段落,我就想上台講。從博士班到現在,我的研究在 DATE、ICCAD、ASP-DAC、ITC、DAC 等國際學術研討會發表了 11 篇論文,其中 ITC、DATE、DAC 都是我親自上台報告。2026 年在美國長灘的 DAC 報告 PG-GNN 時,也在會場認識了來自希臘與法國的研究夥伴。Whenever my research reaches a milestone, I want to present it. Since my Ph.D., my research has appeared in 11 papers at international academic conferences, including DATE, ICCAD, ASP-DAC, ITC and DAC, and I presented in person at ITC, DATE and DAC. Presenting PG-GNN at DAC 2026 in Long Beach, I also met research partners from Greece and France.

新技術我也會帶進課堂。在陽明交大經管所,我連續兩年受邀分享區塊鏈技術、產品旅程與優化。在 AI 工作坊與第二大腦課程裡,學員帶著自己的工作來,一步一步把錄音、筆記與 AI 串成能用的流程。I bring new technology into the classroom too. For two consecutive years, I have been invited to NYCU’s Institute of Business Management to discuss blockchain technology, product development journeys, and optimization. In AI workshops and second-brain courses, participants bring their own work and combine recordings, notes, and AI into practical workflows.

下一次,現場見Meet at the next event

生活化,一步一步,
每段都有小練習。
Everyday examples.
One step at a time.
A short exercise in every section.

這是我的教學原則。讓沒有程式基礎的人,也能帶著成果回去。My teaching principle: help people leave with something that works, even without a programming background.

梁家為在餐桌旁微笑的生活照
Off the clock. Still curious.

故事還在繼續・很高興認識你The story continues · Nice to meet you

下一段故事,
也許可以一起寫。
Perhaps we can build
the next chapter together.

研究合作、產品開發,或是想讓 AI 幫上你的工作與生意。帶著你的問題來,我們從一段對話開始。Whether it is research collaboration, product development, or putting AI to work in your business, bring your question, and let's start with a conversation.

A PORTRAIT OF AARON LIANG回到封面 ↑Back to the cover ↑