Worked with Card Risk Modeling team on Acquisition Premium Gamer model.

• Performed ETL on the waste data and analyzed its statistical patterns, explored novel classification perspectives and identified more precise cutoff, significantly enhanced the accuracy of classification.• Processed data with SQL queries for reporting and API calls from Automated Classification Engine in multiple PostgreSQL databases, which contain more than 30 tables and 1 million rows in total.• Contributed core API services, added input validation, diversified waste code coverage, and increased the robustness of Automated Classification Engine, which avoided millions of dollars loss for customers.• Created Tableau reports, provided customers with data-driven answers, and visualized internal success, increased communication efficiency and maintained consistent satisfaction from 15 customers.

• Analyzed raw data from EPA and Austin Recycle and Reuse Center, implemented recycling bin rules in Lucene query, successfully avoided banned or industrial-only pesticide in the recycle stream.

• Sampled and augmented the thermal images of patients, trained an SSD MobileNet network with 96.19% accuracy on the test set, and enabled recognition of multiple faces with or without face masks in real-time.• Segmented lung area from the images and marked the inflammation area with processed lung mask in OpenCV, significantly increased the diagnosis efficiency by filtering out the potential patients.• Integrated the above modules into the COVID-19 asymptomatic patient surveillance system, successfully identified potential COVID-19 patients in the crowd with no false-negative cases.