- Developed a hospice eligibility prediction model with LightGBM, achieving an AUROC of 0.83 across 1M members, supporting future clinical interventions.- Optimized ER utilization prediction, boosting AUPRC by 10% through advanced feature engineering (e.g., temporal trends, condition severity, medication utilization) and cohort design techniques. - Collaborated extensively with clinicians to design relevant interventions based on model predictions, improving the impact of healthcare solutions. - Conducted member segmentation using clustering techniques (e.g., k-means, hierarchical), identifying distinct groups to prioritize outreach efforts and optimize resource allocation. - Processed ~1.3 billion claims using Apache Spark and SQL on Databricks, cutting processing time by 90%. - Automated MLFlow setup with reusable Python modules, reducing manual setup effort for the team and doubling weekly experiment throughput.

During my internship, I was responsible for creating the infrastructure for monitoring the Emergent Care Transformation model in production. I also was tasked with creating a reusable pipeline for the encoding, imputation, and scaling of new data for integration with the model.- Spearheaded the establishment of a robust monitoring infrastructure for the Emergent Care Transformation model, orchestrating automated monthly metric evaluations to maintain model performance and effectiveness over time.- Utilized SQL and PySpark expertise to compute and analyze 15 physician-recommended metrics from databases containing 30+ tables. - Integrated the Evidently Python package to enable real-time data drift monitoring, saving 5+ hours of labor a week. - Championed the implementation of model fairness calculations, resulting in a 50% reduction in model bias across 5 sensitive features. - Demonstrated exceptional collaboration skills, leading cross-functional engagement and facilitating meetings among 6 team members. - Engineered an end-to-end data preprocessing pipeline to handle encoding, imputation, and scaling of new data; reduced data preprocessing time by 40% and increased model performance by 15%.

I spent nearly two years as a bioinformatic analyst researching post-transcriptional regulation during viral infections. Specifically, I focused on the mechanism by which Dengue virus manipulates host codon optimality to adapt to multiple hosts (i.e. mosquitos and humans) and efficiently propagate in either environment.- Performed large-scale analysis on thousands of viral and human genomes using R and Bioconductor packages, uncovering patterns in gene composition, codon prevalence, and host-pathogen relationships. - Applied the scientific method to design and execute experiments analyzing host-pathogen interactions, ensuring robust and reproducible results. - Authored detailed technical documents, bridging the gap between data-driven insights and healthcare research outcomes. - Delivered more than 20 presentations to audiences as large as 100+ listeners, succinctly conveying intricate concepts and abstract ideas.

I was in charge of leading a lab section of the Introductory to Python course at the University of Arizona. I taught the material, answered questions, held office hours, and graded assignments and exams. - Led curriculum development for Python programming, covering core concepts such as data structures, loops, I/O, functions, and operators.- Mentored students during dedicated office hours, offering guidance and valuable feedback on assignments.- Attained exceptional results with my section, securing the highest scores on the final exam.