• Developed an internal Innovation and Ideas Portal with Spring Framework, leveraging microservices, Spring Security, Eureka, and Kafka to enhance service integration and ensure robust security.

• Streamlined deployment and improved operational efficiency by adopting Docker for containerization and Jenkins for CI/CD pipeline enhancements.

• Increased system reliability by 30% with ELK for centralized logging and proactive monitoring.

• Ensured 99.9% uptime and managed peak loads with AWS (EC2, Auto Scaling) for high availability.

Core-infrastructure

• Improved the server's stability and reliability by refactoring the JRuby application from an outdated application server to Puma.

• Implemented logging feature with Log4j2 and other configurations, such as deflator, that the Puma web server doesn’t provide.

• Created a tool that automatically starts multiple Locust tests with the different arguments in virtual machines and export performance plots using Python and Shell script.

• Constructed a large social network of maritime crime with 53000+ vertices and 2590000+ edges and built exponential random graph models to analyze features of criminals using R.

• Examined the criminals’ and suspects’ collaboration tendencies and behavior patterns by extracted features, such as vertices’ attributes and DID features, and implemented ANOVA using Python.

• Created random graphs of active and criminal networks to calculate the efficiencies and securities and then compared those with real networks’ features using R.

• Enhanced RESTful API’s availability, flexibility, and scalability by restructuring API and building the entire skeleton that meets the OpenAPI specification using Java Vert.x.

• Developed an API gateway instance to deploy API controller verticles, intercept incoming requests, and communicate with those verticles through eventBus.

• Constructed an abstract class that defines the instance’s behaviors and how to invoke the corresponding methods through the arriving message in the API verticle.

• Established a Vert.x instance to consume log data via Kafka queue for follow-up data analysis to gain valuable perception from our clients, then construct a more insightful strategy.

• Decreased up to 90% of time and cost from the manual key-in method by developing a questionnaire recognition system using Django. The system scans PDF files, loads subject information via QR codes and locates each question, and saves the results to the database.

• Started a rule-based model from feature engineering results to detect whether the check box is checked and collected the handwriting section for the user to identify.

• Instituted a prototype of a real-time online-experimental platform with event listeners using Google Firebase and Python.