Wen-Yuh (Ken) Su Email & Phone Number

Ottawa, Ontario, Canada

Taipei City, Taiwan

Azure Stack Team (Python, Go, Java, Kubernetes, Terraform, Azure Stack Hub, MySQL)• Automated and refined the public and private keys, and service JWT tokens generation flow in the private cloud while porting the Service Platform and UI Container app to avoid manual works and duplicate efforts• Implemented the active update feature for the Azure Stack Hub environment to renew filters, rules, and models• Re-designed ETL Filter, API Server, and Log Receiver in SAE to adapt the Azure Stack Hub environment by changingfrom NoSQL to MySQL and modifying the Azure SDK in various programming languages within a short periodThird-Party Integration Team (Python, Flask, Go, PostgreSQL, Restful API, Kubernetes, KEDA, RabbitMQ)• Improved a server-agent communication application for customers to set suspicious object sharing configurations onthe SaaS platform by an asynchronous message queue consuming method to decrease the waiting time• Designed the third-party threat intelligence application structure with Kubernetes and RabbitMQ which can processcyber threat intelligence feeds from servers with high throughput and reducing the cost of maintenance• Applied KEDA to build auto-scaling microservices that improve efficiency and save 80 percent of processing time

Mountain View, California

• Applied embedding similarity features by training word2vec to improve disambiguation of names for US cities and counties with F1 measure of 78.48 percent• Implemented XGboost model to recommend the news to users living in the related area with accuracy of 90 percent• Built the data pipeline for the push-notification system with extracted time and action with predicted event type byusing SUTime and the dependency parser and resulted in increasing CTR with 10 percent on A/B Test

Greater Chicago Area

• Implemented Hierarchical LSTM which combined the word embedding sequence and the clause embedding to achieve high accuracy of 78.6 percent on classification task with 4 labels on Twitter HPV-related articles• Compared effectiveness of the semantic representation by applying latent semantic analysis, Word2Vec, and FastText on the different level of words in HPV-vaccine-related data

Taipei City, Taiwan

• Emphasized risk detecting for word-level financial reports by training a self-attention LSTM model on the collection of financial reports with financial sentiment phrases, resulting in high accuracy of 88 percent (ICASSP’20, 2nd author)• Built a labeling system by using Flask framework, JavaScript and MongoDB for users to add annotations of multiword expressions for financial reports with visualization which increased the efficiency of labeling• Visualized the financial statement by highlighting the strong words which the deep learning model learned

Taiwan Taipei

Teaching Assistance

Taipei City, Taiwan

• Applied Lambdamart model on financial data such as financial behavior of users, truncation record, and customer profiles to find important features• Implemented the Deep-Wide model to predicted the average variation in the amount of credit card transaction of customers and the features from XGBoost with a high accuracy

Taipei City, Taiwan

• Implemented university dynamic map on NCCU App, and promoted app usage, increasing the number of totalusers from 150 to 10,000• Improved a questionnaire system with RWD interface for all university users, successfully increasing its usage

Master'S Degree, Computer Science

Bachelor'S Degree, Computer Science

Activities and Societies: Chia-yun Senior High School Alumni Society