許應良 · Joe HsuJoe Hsu · Yingliang Hsu

演算法工程師｜AI / Computer Vision / AutomationAI / Computer Vision / Automation Engineer

joe0827joe@yahoo.com.tw · 新北市, 台灣 · Portfolio · GitHub joe0827joe@yahoo.com.tw · New Taipei City, Taiwan · Portfolio · GitHub

專業摘要Professional Summary

北科大自動化科技碩士，專長深度學習、影像辨識與工業 AOI 系統落地。具備從模型訓練、推論優化到跨部門工程驗證的完整實作經驗，能把演算法轉成可量化、可部署的解決方案。Automation Technology M.S. (NTUT) specializing in deep learning, computer vision, and industrial AOI delivery. Experienced from model training and inference optimization to cross-functional validation—turning algorithms into measurable, deployable systems.

關鍵成果Key Results

94.44% AOI 召回率AOI Recall7% PUE 降低PUE Reduction15% LED 體積縮小LED Volume Cut10% 亮度提升Brightness Gain

專業技能Technical Skills

LanguagesLanguagesPython · C/C++ · C# · JavaScript · VBA

AI / CVAI / CVPyTorch · TensorFlow · OpenCV · Scikit-learn · NumPy · Pandas

DeploymentDeploymentONNX · TensorRT · OpenVINO · Docker · Linux

EngineeringEngineeringGit · GitHub Actions · Pytest · CI Workflows

工作經歷Professional Experience

軟體工程師Software Engineer

2023/07 – 至今2023/07 – Present

易豐購企業社Yifong Enterprise Co., Ltd. · 新北市New Taipei City

開發 Python 自動化與跨平台資料轉譯，涵蓋 API Schema Mapping、SKU 屬性整理與非結構化資料轉換Built Python automation and cross-platform data translation for API schema mapping, SKU attribute processing, and unstructured data conversion
依 ROI、流量成本與商品條件設計規則型邏輯，支援自動化篩選、排序與資源分配Designed rule-based logic using ROI, traffic cost, and product conditions to support automated filtering, ranking, and resource allocation
導入 Git、例外處理與日誌追蹤，提升無人值守腳本的穩定性與可維護性Improved unattended script reliability with Git version control, exception handling, and log tracking

車用 LED 元件研發工程師（研替）Automotive LED R&D Engineer (Alternative Service)

2022/01 – 2023/062022/01 – 2023/06

億光電子工業股份有限公司Everlight Electronics Co., Ltd. · 新北市New Taipei City

導入 Top Contact 散熱與新晶片架構，達成體積縮小 15%、亮度提升 10%，並完成車用信賴性驗證Implemented top-contact thermal path and new die design, achieving 15% volume reduction and 10% brightness gain while completing automotive reliability validation
依 APQP / AEC-Q 管控專案節點，協作品保與製程縮短驗證時程Managed APQP/AEC-Q gates and coordinated QA/process partners to shorten validation cycles

CNC 技術人員（實習）CNC Technician (Intern)

2016/08 – 2018/072016/08 – 2018/07

群光電能科技股份有限公司Chicony Power Technology Co., Ltd. · 新北市New Taipei City

開發深孔加工治具抑制震動與飄移，治具成本降低 20%Developed deep-bore fixtures to suppress vibration and drift, cutting jig cost by 20%
與研發進行 DFM 評估並優化加工路徑，誤差率下降 30%Partnered with R&D on DFM reviews and tool-path optimization, lowering error rate by 30%

精選專案Selected Projects

PCB 混合動力瑕疵檢測系統PCB Hybrid Defect Detection System（主打） (Featured)

C++ · Python · OpenCV · OpenVINO · Deep Learning · 專案連結Link

整合規則式 ROI 定位與深度學習分類，於工業 IPC 環境達成低延遲、高精度檢測，無需高階 GPUCombined rule-based ROI localization with DL classification for low-latency, high-accuracy inspection on industrial IPCs without high-end GPUs
CV Engine 於 10ms 內完成 ROI 定位，節省約 80% AI 推論算力；OpenVINO INT8 優化 Intel CPU 推論CV engine localized ROI in <10ms, cutting ~80% AI inference cost; OpenVINO INT8 optimized Intel CPU inference
DL Engine 支援 6 類疵點分類；導入 Pytest 與 CI 確保演算法迭代穩定DL engine classifies 6 defect types; Pytest + CI keep algorithm iterations stable

資料中心冷卻節能優化Data Center Cooling Optimization

Python · DE-MRBFNN · NSGA3

以 DE-MRBFNN + NSGA3 建立氣冷單元式資料中心節能模型，實驗環境降低 7% PUEBuilt DE-MRBFNN + NSGA3 energy model for an air-cooled modular data center, cutting PUE by 7% in experiments
將 CFD 模擬耗時由數小時降至秒級，並以研討會論文發表成果Reduced CFD simulation runtime from hours to seconds; published in a peer-reviewed symposium

Anime4K C# 影像強化工具Anime4K C# Image Enhancer

C# · .NET 9 · WinForms · 專案連結Link

實作 Anime4K 核心演算法與 WinForms 即時參數調校介面，強化動漫風格邊緣與質感Implemented Anime4K core algorithm with WinForms live tuning for anime-style edge and texture enhancement
以 LockBits 直接操作記憶體優化 CPU 效能，支援 Scale／Strength／迭代參數即時調整Used LockBits for CPU-efficient memory access; real-time Scale/Strength/iteration controls

學歷Education

自動化科技研究所碩士M.S., Automation Technology

2019/08 – 2021/092019/08 – 2021/09

國立臺北科技大學National Taipei University of Technology

氣冷資料中心多目標節能模型（DE-MRBFNN + NSGA3），PUE 降低 7%，研討會論文發表Multi-objective data-center energy model (DE-MRBFNN + NSGA3); 7% PUE reduction; symposium paper

機械工程系學士B.S., Mechanical Engineering

2015/09 – 2019/062015/09 – 2019/06