Experiences

2025/07/07~Now,  Senior Manager (協理)
Technology & Transformation - Cyber, Taipei, Taiwan

TBD

2023/12/11~2024/07/05,  Manager / Department Head (部經理)
Telecommunications Service Provider, Taipei, Taiwan 想不開？？

Responsible for developing information systems tailored for AI and cybersecurity, spanning machine learning, deep learning, and other advanced computational techniques. Projects include dark web threat intelligence platforms, ASR error correction, accounting document OCR, large language & multimodal models, and multi-agent workflow automation systems.

開發專為人工智慧與資安領域量身打造的資訊系統，涵蓋機器學習、深度學習及其他先進運算技術。專案包括暗網威脅情報平台、自動語音辨識錯誤校正、會計憑證 OCR、大型語言與多模態模型，以及多智能體與流程自動化系統。

• 威脅情報 / Threat Intelligence：暗網監控與分析平台 / Dark web monitoring and threat intelligence platform.

• ASR 錯誤校正 / ASR Error Correction：提升語音辨識準確度 / Improved ASR accuracy.

• 會計憑證 OCR / Accounting OCR：自動化財務單據辨識 / Automated document recognition.

• 大型語言與多模態模型 / LLM & Multimodal Models：應用於多元商業場景 / Applied to diverse business use cases.

• 多智能體與工作流程系統 / Multi-agent & Workflow Systems：流程自動化設計 / Workflow automation design.

2022/07/04~2023/12/11,   Senior AI Technical Architect (資深AI技術架構師)
Digital, Data and Technology, Cathay Financial Holdings, TW
數位數據暨科技發展中心 (數數發)，國泰金融控股股份有限公司，台灣

AI architecture and solution development focusing on natural language processing, OCR, and speech recognition.
人工智慧架構與解決方案開發，專注於自然語言處理、光學字元辨識與語音識別。

• 命名實體識別與事件擷取 / NER & Event Extraction：醫療診斷書分析與日期擷取 / Medical diagnosis analysis & time extraction.

• OCR / OCR：文本、表格偵測與關鍵資訊擷取 / Text, table detection, and key information extraction.

• 語音識別與去噪 / Speech Recognition & Noise Reduction：遠距投保與客服語音質檢工具 / QA tools for remote insurance and customer service speech interactions.

• 受 bbc.com 專訪，Taiwan's front-line battle against mobile phone fraud @ bbc.com 報導。

2015/12/28-2021/03/21, Senior Technical Expert and Team Lead (資深技術專家 / 資訊安全與人工智慧)
Leopard Mobile (Cheetah Mobile TW Agency), Taiwan (台灣雪豹科技 / 獵豹移動台灣總代理，台灣)

Led the design and deployment of machine learning (ML) and deep learning (DL) products, including algorithm development, requirements analysis, architecture design, proprietary dataset creation, model quantization, and optimization, achieving millisecond-level inference latency while reducing model size and enhancing precision.
In cybersecurity, specialized in malware detection, blockchain vulnerability assessment, and cryptocurrency ranking, with research presented at international security conferences such as Defcon, RuxCon, OWASP Appsec, and HITCON.
AI expertise covered speech processing, NLP, and computer vision, with a focus on speech separation, enhancement, speaker recognition, and speech recognition, integrating NER, classification, similarity analysis, and machine reading comprehension to enhance AI assistants’ ability to understand and respond to human intent. 

負責規劃與建置機器學習（ML）與深度學習（DL）產品，涵蓋演算法設計、需求分析、技術架構、數據集構建、模型量化與優化，並將模型推論速度優化至毫秒級，同時減少模型大小與提升精準度。 在資安領域，專注於惡意程式檢測、區塊鏈漏洞分析與加密貨幣評級，並於 Defcon, RuxCon, OWASP Appsec, HITCON 等國際安全會議發表研究成果。 人工智慧應用涵蓋語音處理、自然語言處理與電腦視覺，專精於語音分離、語音增強、聲紋辨識與語音識別，並結合 NER、文本分類、相似度分析與機器閱讀理解，讓 AI 機器人能更精確理解與回應人類意圖。

• 智慧服務型機器人與 AI 助理 / Intelligent Service Robots & AI Assistants
  Built ML/DL applications for image, NLP, and speech processing to enhance interactive experiences.
  開發影像、自然語言處理與語音處理應用，全面提升互動體驗。

• 惡意廣告與騷擾電話識別 / Malicious Ad & Spam Call Detection
  Applied DL models to detect malicious messages, significantly improving accuracy.
  採用深度學習模型有效偵測惡意訊息，顯著提升準確度。

• 聲紋識別與語音去噪 / Speaker Recognition & Speech Denoising
  Integrated SincNet, Kaldi, and Facebook Denoiser, optimized with aishell and ESC-50 datasets, achieving 0.94 speaker recognition accuracy for service robots.
  整合 SincNet、Kaldi 與 Facebook Denoiser，並利用 aishell、ESC-50 數據集優化，使聲紋辨識準確率達 0.94，應用於服務型機器人。

• 語音分割 / Speech Segmentation
  Enhanced speech separation with RNN/CNN architectures, self-supervised models, and multi-resolution feature downsampling.
  結合 RNN/CNN 結構、自監督模型與多解析度特徵下採樣，提升語音分離效果。

• 自然語言處理與機器閱讀理解 / NLP & Machine Reading Comprehension
  Optimized BERT and developed custom annotation tools, supporting SQuAD 1.0/2.0 and DRCD to improve semantic understanding accuracy.
  優化 BERT 模型並建立數據標註工具，支援 SQuAD 1.0/2.0 與 DRCD，提升語義理解精準度。

• 文本糾錯與相似度分析 / Text Correction & Similarity Analysis
  Leveraged ConvSeq2Seq and BERT to improve inference speed and accuracy for semantic identification.
  基於 ConvSeq2Seq 與 BERT，提升推論速度與準確率，加強語意識別。

• 命名實體識別與文本分類 / NER & Text Classification
  Integrated pinyin-based classification for marketplace data, improving recognition accuracy and containerized API inference performance.
  整合拼音與分類機制，應用於商場數據，提升識別精準度與 API 推論效率。

• 支援台灣防疫技術，提升聲紋辨識正確率以應用於防疫系統 (科技抗疫: 提高聲紋辨識正確率，添防疫新利器)。
  Enhanced speaker recognition accuracy for Taiwan’s pandemic prevention technologies.

• 作為 NVIDIA 開發者，應用 AI 提升行動應用程式通知的頻率與品質。
  As an NVIDIA Developer, applied AI to improve the frequency and quality of mobile app notifications.

• 受媒體專訪，列入 Career「十大熱門人才」系列報導。
  Featured in Career’s “Top 10 In-demand Talents” series.

2014/07/07-2015/12/14, Senior Security Engineer (資深資安工程師)
Verint Systems (Taiwan), Taipei, Taiwan (台灣威瑞特系統)

Developed Android security architecture, specializing in malicious program analysis and detection with in-depth research in cybersecurity. Key achievement: APKProbe — a hybrid Android malware probing system combining semantic-based static analysis to bypass code obfuscation and physical cluster-based dynamic analysis to evade anti-sandboxing techniques. Additional responsibilities included delivering custom training, designing solution architectures, conducting cybersecurity research, providing product education and client support, monitoring security equipment, performing incident analysis and response, and conducting digital forensics.

負責開發 Android 安全架構，專注於惡意程式分析與偵測，並深入資安領域研究。核心成果包括 APKProbe：混合式 Android 惡意程式探測系統，結合基於語義的靜態分析以破解代碼混淆，並運用實體叢集機器的動態分析技術以規避反沙箱檢測。 其他工作內容涵蓋客製化教育訓練、資安解決方案架構設計與研究、資安產品培訓與客戶維護、資安設備監控、事件分析與應變，以及數位鑑識。

• APKProbe Malware Probing System
  Created a hybrid probing framework for Android applications integrating static and dynamic analysis to improve malware detection accuracy.
  創建針對 Android 應用程式的混合式探測架構，整合靜態與動態分析技術，提升惡意程式偵測準確度。

• 資安教育與培訓 / Cybersecurity Education & Training
  Designed and delivered enterprise-level security training to strengthen client protection capabilities.
  設計並提供企業級資安教育課程，提升客戶端防護能力。

• 事件分析與數位鑑識 / Incident Analysis & Digital Forensics
  Monitored security systems, investigated incidents, and performed digital forensic analysis.
  監控資安設備、分析異常事件並執行數位鑑識作業。

2014/01/20-2014/07/04, Senior Security Engineer (資深資安工程師)
Acer Cyber Security, Taipei, Taiwan (安碁資訊股份有限公司)

Developed Acer Cyber Evidence, a cluster-based digital forensics toolkit integrated with Arcsight to enhance threat detection and evidence collection capabilities. Focused on network threat analysis, malware research, and digital forensics technology development. Responsibilities included delivering custom training, product maintenance, solution planning, monitoring security systems, incident analysis and response, and performing digital forensics operations.

開發 Acer Cyber Evidence 數位鑑識叢集服務工具，並與 Arcsight 整合，以強化威脅偵測與證據收集能力。專注於網路威脅分析、惡意程式研究與數位鑑識技術開發，負責客製化教育訓練、產品維護、解決方案規劃，以及資安設備監控、事件分析與應變、數位鑑識作業。 

• Acer Cyber Evidence 開發 / Acer Cyber Evidence Development
  Designed and implemented a cluster-based digital forensics service integrated with Arcsight, improving threat detection efficiency and forensic capability.
  設計與實作數位鑑識叢集服務，結合 Arcsight 提升威脅偵測效率與取證能力。

• 網路威脅與惡意程式分析 / Network Threat & Malware Analysis
  Developed tools for threat analysis and malware research, enhancing forensic accuracy and response time.
  開發威脅分析與惡意程式研究工具，提升鑑識準確性與反應速度。

• 資安教育與解決方案規劃 / Cybersecurity Training & Solution Design
  Delivered enterprise-level security training, product maintenance, and solution planning.
  提供企業資安教育訓練、產品維護與解決方案規劃。

2008/11-2011/09/28, Assistant Researcher (專案助理研究員 / 研發替代役)
National Center for High-Performance Computing, Tainan, Taiwan (國家高速網路與計算中心)

Integrated Diskless Remote Boot in Linux (DRBL) to develop TWMAN, an open-source automated malware behavior analysis toolkit for Windows platforms. Responsible for cybersecurity research, developing malware analysis techniques, and managing operations of the Southern Taiwan Academic Network. Additional duties included delivering custom training, conducting security solution research, and providing technical support.

整合 Linux 無碟遠端啟動（DRBL） 技術，開發 TWMAN（抬丸郎） —— 基於 Windows 平台的開源自動化惡意程式行為分析工具。負責資安研究與惡意程式分析技術開發，並管理南部學術網路的維運。工作內容包含客製化教育訓練、資安解決方案研究與技術支援。

• TWMAN 惡意程式分析工具 / TWMAN Malware Analysis Toolkit
  Developed an open-source automated malware behavior analysis toolkit for Windows, integrating DRBL to improve analysis efficiency and scalability.
  開發基於 Windows 的開源自動化惡意程式行為分析工具，整合 DRBL 提升分析效率與可擴充性。

• 惡意程式分析與資安研究 / Malware Analysis & Security Research
  Developed behavior analysis and forensic techniques to strengthen cybersecurity capabilities.
  研發行為分析與鑑識技術，強化資安防護能力。

• 南部學術網路維運 / Southern Taiwan Academic Network Operations
  Managed network infrastructure ensuring high availability and security.
  管理網路基礎架構並確保高可用性與安全性。