Chen-Wei Sung Email & Phone Number

台灣 臺北市 台北

► Enterprise Customer Insights• 將保單從招攬、核保、再保與理賠等資料底層進行串接，建構以客戶為中心的 Data Model。• 從客戶細到每張保單，梳理貢獻計算方法，以衡量每個客戶對於公司的貢獻程度，並針對各類客群提供經營方針之建議。► 損失樣態分析• 利用開源 LLM 模型建構資訊擷取的 Pipeline, 讓理賠人員能透過簡單的方式設定 Prompt 去擷取理賠文件中的重要資訊，協助損失樣態分析。#Prompt Engineering #Chatbot

台灣 臺北市

► Marketing Tags• [PM] Collaborated with 6+ product departments to establish the tagging system, improving the engagements from digital marketing channels.• [PM] Leveraging interviews, workshops, and presentations to ensures the product enhances the efficiency of marketing material creation.• [Data] Implemented user tagging strategies utilizing marketing interaction frequency, text analysis of marketing materials, and campaign timeliness.• [Data] Designed A/B testing strategies to evaluate the effectiveness of our tagging system.► Product Recommendation Optimization• [PM] Led a team of 5+ members, including data analyst and ML engineer, to ensure effective communication and coordination among team members for successful job delivery.• [Data] Evaluated model's ranking performance across customers and products to verify its ability to reduce the number of ongoing models.• [Data] Probability calibration: Improves model ranking performance while providing clearer output interpretation.► Costco customer retention analysis• [BA] Built an analytical architecture using CLV and churn risk to identify the most valuable customers for retention in the company.• [BA] Worked closely with product departments to develop marketing strategies and budget analysis for different customer segments.

台灣 新北市

As the project manager for the recommendation system for 3 different news websites, including the United Daily News, which is the 2nd most visited news website in Taiwan, my role involves:► Project Management• Working with stakeholders from each news platform to understand their specific needs and goals for the recommendation system.• Collaborating with cross-functional teams, including editors, UX/UI designers, and backend developers, to ensure seamless integration of the recommendation system into the websites.► Recommendation System Development • Designing and building the end-to-end architecture of a recommendation system on GCP, covering data integration, storage, model training, and deployment. - data pipeline: BigQuery, Dataflow, and Pub/Sub - model deployment: Cloud Run, Functions, and BigTable - maintenance: Airflow, and Data Studio• Developing personalized recommendation algorithms using collaborative-filtering, keyword extraction, news-embedding, and ANN search to provide tailored content.• Compared to the original algorithms, CTR shows a significant 30% increase in each position.

台灣 新北市

► Precision marketing• Working closely with the sales team to define Ads categories and developing a classification model based on Ads description to automate daily labeling tasks.• Integrating user behavior data from news and e-commerce sites to build click prediction model, leading to a 50% increase in display-ads CTR.• Utilizing user-embedding obtained from recommendation model to expand the pool of seed users and fulfill the impression requirements of advertisers.

Master'S Degree, Finance

Activities and Societies: 【 玉山銀行大數據競賽 - 決賽 】 利用信用卡消費與網頁瀏覽數據了解顧客行為模式，結合客戶於國外刷卡與外幣投資兩大外匯操作行為，建構數位化旅遊投資基金，並針對不同年齡層的客群差異，提供多樣化的行銷策略規劃，提升整體外匯的換匯.

Bachelor'S Degree, Information Management