Prys Wang

• Helped with grading assignments and tests, providing constructive feedback to students based on results.
• Tutored struggling students individually and in small groups to reinforce learning concepts.
• Worked with teacher to optimize learning plans based on student test grades and evaluations.
• Taught courses under guidance of senior instructors.

• Cleaned, processed, and analyzed large-scale customer behavior and sales datasets (5M+ records) to identify purchasing trends and seasonal demand patterns.
• Developed automated reporting dashboards in Tableau for sales and marketing teams, reducing manual report preparation time by 60%.
• Built predictive models in Python (XGBoost, Random Forest) to forecast monthly sales with a mean absolute error of less than 8%.
• Collaborated with cross-functional teams to design data-driven strategies, improving campaign ROI by 15%.

• Conducted data cleaning, anomaly detection, and pattern recognition to identify abnormal operational units in large-scale wind farms using Supervisory Control and Data Acquisition (SCADA) and Condition Monitoring System (CMS) datasets.
• Maintained and continuously updated algorithmic models deployed on the big data platform to ensure stable and accurate wind turbine performance monitoring.
• Developed and validated a CMS-based blade vibration fault diagnosis model to assess structural integrity and operational risks of turbine blades.
• Researched and implemented machine learning and deep learning techniques (e.g., random forest, CNNs) for predictive maintenance and fault classification in wind turbine systems.
• Collaborated with cross-functional teams to interpret results and improve the reliability and energy efficiency of wind energy systems.

Project: Credit Card Customer Profiling & Loan Default Prediction

• Analyzed large-scale credit card transaction and customer demographic data to build user personas, segmenting customers into behavioral clusters using K-Means and hierarchical clustering.
• Engineered 50+ behavioral and financial features (e.g., spending patterns, repayment history, credit utilization rate) to enhance model performance.
• Developed an XGBoost-based loan default prediction model, achieving an AUC of 0.86 and reducing Type II error by 15% compared to baseline logistic regression.
• Designed an interactive Tableau dashboard to visualize customer segments and default risk levels, supporting targeted marketing and risk control strategies.

Project: Advertising Investment Analysis & Sales Forecasting

• Integrated and cleaned advertising expenditure, sales, and seasonal campaign data from SQL databases.
• Conducted correlation and regression analysis to quantify the ROI of different ad channels, identifying a 22% higher return from digital campaigns versus traditional media.
• Built a Prophet-based time series forecasting model to predict monthly sales, improving forecast accuracy by 12% over previous methods.
• Presented findings to the marketing team through an interactive Tableau dashboard, enabling data-driven budget reallocation.