• Led the improvement and optimization of the Bayesian Hierarchical Integrative Model for multi-omics data, integrating over 3,000 miRNA-mRNA pairs and applying biological domain expertise to pinpoint potential biomarkers for hepatocellular carcinoma (HCC) and advanced predictive oncology and genomic analytics by associating genomic pairs with disease traits. Additionally, proposed a biological reference network as a novel method of biological validation for comparative analysis across diverse methodologies.• Refined existing Bayesian calculation R code for efficiency and am co-developing a package for Hierarchical Integrated Bayesian miRNA-mRNA analysis with biological validation, publication in progress.

- Conducted a comparative analysis of autoencoders and traditional dimension reduction methods like PCA, t-SNE, and UMAP, aiming to provide meaningful lower-dimensional representations of high-dimensional multi-omics data while retaining critical biological insights, especially in miRNA-mRNA associations.- Utilized Bayesian Neural Networks to predict significant miRNA-mRNA pairs from Variational Autoencoder compressed genomic data, aiming to identify potential biomarkers for Hepatocellular Carcinoma (HCC), enhancing the precision and understanding of genomic associations in cancer biology.

- Mining metrics like each carrier’s average shipping time and delay in shipments from the company’s monthly shipment data in MySQL, analyzed and visualized these data in Excel and Tableau. Individually generated reports of the current shipping qualities and presented in department meetings- Developed a shipment qualification model based on over 100,000 user monthly feedback for the logistic team. Utilized principal component analysis to reduce the dimension of the dataset and perform feature selection to reduce multicollinearity for the later linear regression to predict the shipment on time rate for each order. Improved the accuracy of delivery time prediction by 13% from the original 63% baseline and consumer satisfaction by 10% from the original 68%.

- Extracted over 200,000 weekly user activities data from the company’s SQL database, generated user metrics such as average training time, premium account subscription rate, and average course completion rate and cleaned up the datasets with NumPy and pandas. Visualized the weekly changes with Matplotlib and report in department’s weekly meeting.- Performed A/B testing with multiple metrics, interpreted significance of the results with F statistic test and linear regression. Led to the implementation of a new UI feature in mobile app that increased 9% of user engagement rate