At Mary Wade Home, I’m conducting a multi-year actual-to-expected experience study that brings together census, staffing, budget, and financial data. I’m also developing evidence-based budget-allocation scenarios using resident surveys, service expenditures, and external benchmarks.
On my 30-day readmission prediction project, I built a BigQuery pipeline integrating structured EHR data and clinical notes, then developed and validated a stacked Clinical-Longformer model. It improved mean AUROC from 0.711 to 0.744 across five patient-level splits.
My other projects include a quantitative research platform for U.S. P&C and reinsurance equities, and actuarial analysis of cluster randomized trial data. I’ve applied methods including XGBoost, modified Poisson regression, GLMMs, Monte Carlo simulation, and Bayesian inference.
