Research Associate

Built a forecasting model that gives triage teams an early estimate of how long incoming patients may wait before being seen.

Hospitals often lack an accurate real-time picture of queue length, so by merging triage, staffing, and census data we developed a gradient-boosted model that delivered wait-time estimates within eight minutes. Nurses reported improved scheduling confidence and clearer communication with patients, and their feedback guided refinements to the user interface. Development relied on pandas for feature engineering, SHAP for model explainability, and a lightweight Flask application for demonstration.