Customer Detector• To assist fundraising managers in reducing operational costs and optimizing the donor acquisition process, built an end-to-end Donor Detection system using binary classification models.• Acquired an online dataset of 6.8 million donors; performed exploratory data analysis (EDA), data cleaning, preprocessing, feature engineering, and feature selection; addressed the challenge of big data analysis by utilizing a cloud computing platform; trained and fine-tuned machine learning models (Logistic Regression, Random Forest, XGBoost) to predict an individual's likelihood of donating to the organization.• Achieved a ROC-AUC of .93, increased precision from .56 to .90, potentially adding $25 million in annual donations. Wrapped the best-performing Random Forest model as a Flask App and deployed it on GCP.Sales Forecasting• To support political fundraising managers in effective financial planning and resource allocation, developed a Donation Forecasting model to predict monthly donations using a 10 GB time series dataset.• Collected and organized data, implemented label and frequency encoding, conducted EDA, and generated new features from timestamps. Built and fine-tuned Linear Regression and ARIMA models.• Based on MAE, MSE, and RMSE metrics, the Linear Regression model outperformed, improving the predictive power by 37% compared to the baseline model.

Received the Outstanding Chair Award in 2020 and 2022.Optimizing Leadership Development with Machine Learning• To enhance leadership training and employee satisfaction, defined six Servant Leadership components as features and Employee Satisfaction as the target. • Collected data using online surveys from an E-commerce organization, addressed missing values and outliers, and applied diverse machine learning models (Linear Regression, Polynomial Regression, Lasso, Ridge, Random Forest, and XGBoost).• Considering the computational cost, interpretability, and metrics such as Mean Absolute Error (MAE), Mean Squared Error (MSE), and R-squared (R²), the Random Forest model emerged as the top performer. Enhanced predictive accuracy by 78% over the baseline model. Provided actionable insights to the organization to refine leadership training programs.Segment Employees using Machine Learning• To reduce technostress and improve well-being among employees at U.S. higher education institutions, segmented employees based on levels of Technostress, including techno-overload, invasion, complexity, insecurity, and uncertainty, along with demographic variables.• Utilized data collection and preprocessing techniques, applied Principal Component Analysis (PCA) for dimensionality reduction, trained, evaluated, and tuned machine learning clustering algorithms (K-means, Hierarchical Clustering, and DBSCAN) to identify four employee clusters based on their technostress profiles.• Provided data-driven insights on customized training and support programs to mitigate technostress and enhance employee satisfaction in higher education

Analysis of Individual Donation Intention• To forecast individual donation tendencies toward collegiate' eco-friendly football stadium projects, designed a survey, and collected data through a mix of paper-based and online methods.• Cleaned data and performed EDA, conducted statistical analyses in SPSS, including Hierarchical Regression, Pearson Correlation, ANOVA, and Student's t-test.• Improved the predictive power by 64% compared to the baseline model. Offered stakeholders data-driven insights into opportunities and challenges for increasing donations. Presented the research findings at multiple national conferences and published them in peer-reviewed journals IJSMM and IJEMR.

Florida Bicycle Safety ProgramRole: Searching reports and journal articles for literature review, conducting data collection and data analysis, and writing reports.