• Conducted extensive exploratory data analysis and feature engineering to understand the impact of customer demographics, service subscriptions, and charges analysis on churn behavior. • Developed a classification model with an 86% recall (0.38 improvement over the baseline model) and recommended targeted interventions to enhance customer retention and lifetime value.

• Conducted Apple Health data analysis and developed a dashboard to visualize events. The application, built using Python with Streamlit, offers interactive plots and insightful metrics.

• Utilized machine learning models, observational data and doubly robust estimation in causal inference to estimate the effects of covariates and treatments. • Proposed a tree-based treatment assignment policy to increase rewards and a method of policy interpretation to demonstrate decision rules.

• Developed a multi-class image classification model (EfficientNet) achieving 84.78% accuracy. The model has been deployed to production, resulting in an 80% reduction in manual classification. • Performed exploratory data analysis on anomalies and simulated defect data to augment the training dataset, which was validated by domain experts.

• Constructed a health indicator from vibration signals via feature engineering, hypothesis testing and PCA. • Built a statistical model (exponential degradation model) with Bayes’ theorem to estimate RUL for predictive maintenance, preventing severe failures and saving ~NT$ 1 million cost.

• Developed a movie recommendation system based on Graph Neural Network (GNN), singular value decomposition and k-NN to capture the relationships between movie ratings and users. • Incorporated explainable AI (LIME) to provide insights into the decision-making process.

• Introduced an open-sourced image annotation tool to tsmc and modified corresponding authentication and authorization mechanism to meet data privacy requirements. • Partnered with API team and cloud service team to provide APIs and container-based web services.