Working under Professor Ran Xiao on an NLP based study

• Identified key predictors for artery blockages in NSTE-ACS patients by performing advanced statistical tests including t-test, z-test and ANOVA on an EHR and ECG database of 6k patients, helping doctors prioritize treatments and contributing to two research papers.• Improved data quality by 30% using Python for extensive cleaning of EHR, ECG data, and clinical notes, ensuring high-quality inputs for predictive modeling for improved clinical outcomes.• Applied Latent Dirichlet Allocation (LDA) in R to extract 8 key topics from clinical notes identifying word patterns linked to higher artery occlusion risks, leading to more accurate predictions.• Built an ensemble model combining multiple algorithms (e.g., Random Forest, XGBoost, Logistic Regression) using document topic probabilities from patient notes to predict artery occlusion, achieving 71% accuracy and improving the prediction of occlusion risk in clinical settings.• Increased model accuracy by 10% through logistic regression, optimizing log-odds ratios for better prediction of occlusion likelihood.• Performed stratified analysis by gender and age, identifying them as key predictors of occlusion for more personalized clinical insights.• Analyzed ECG data from firefighters using Frontier X2 wearables, enhancing a T-CNN-based heart monitoring model.• Prepared an FDA 510k database for Philips by extracting ECG data from 40,000 patients using proprietary software, and applying Python (Pandas) for data cleaning and deidentification to ensure compliance for diagnostic algorithm validation.

• Designed and implemented an optimized PostgreSQL database for managing large-scale biomechanics data, successfully reducing query time by 30%.• Assisted in developing an efficient data processing pipeline for clinical report generation using Python and MATLAB, boosting operational efficiency by 25%.• Managed the input and reporting of client data for clinical and research purposes, ensuring HIPAA compliance and utilizing EPIC software for accurate data handling and secure management.

• Conducted comprehensive data cleaning and preprocessing using Excel and Python (Pandas), reducing data irregularities by 20% and revealing critical patterns in client datasets.• Designed and developed interactive dashboards and visualizations in Tableau, increasing client data insight and improving decision-making efficiency by 30%.