・Collaborated with UX researchers to build a user persona and implemented customer segmentation via cluster analysis. Improved efficiency in the process of finding suitable interviewees and reduced probabilities of selection bias by providing a customer classification model. ・Developed tools and services such as modeling, dashboard automation, and business metric analysis to validate assumptions quickly and drive informed decision aids and recommend solutions, further improving the company’s enterprise product.

・Built a distributed data warehouse and ETL pipeline that can integrate big data from semiconductor equipment in real-time and provided a query system on a millisecond level. ・Developed statistical models to detect the outliers in manufacturing data by IQR and compared the differences in distribution between 2 datasets with EMD Algorithm.

Provide AI solutions, design, develop algorithms, and deploy models involving fields such as computer vision, NLP, and knowledge graphs. Research innovative and cutting-edge technologies that benefit the team’s project goals. ・Directed the development team to implement a complex natural language QA system based on the knowledge graph. I was responsible for ML algorithms part and made 3 presentations to share research results and insights. The participants included 77 individuals and 8 teams. ・Implemented an English-Chinese translation model using Transformer and leveraged the new model process by introducing techniques and results to the team. ・Conducted experiments to compare the performance of Dongle (Kneron and Intel), optimized the ONNX and .h5 models (pruning and constant), and successfully reduced the computing time by 55%. ・Refactored a backend system to an ORM structure by using Sails.js, reduced 20% of code size, and improved maintainability. ・Designed and provided deep learning models for AOI image classification, achieved false positive rate below 5% and false negative rate below 0.5%. ・Developed a Python desktop GUI application (PySide2), which can asynchronously communicate with the backend system, and achieve high-quality reliability. (more than 99% of operating time)