-
Predicting 30-Day Readmission After Cardiac Surgery Using MIMIC-IV Dataset (R Shiny dashboard)
- Built a data pipeline from Physionet BigQuery EHR data, combining hospital admission, ICU vitals, labs, and EKG tables across 500 million rows and 50 GB.
- Ran and evaluated prediction models in R, achieving ROC AUC values around 0.65.
- Developed a Shiny exploratory dashboard for cohort EDA, hyperparameter tuning, admission-discharge-transfer analysis, and historical ICU vitals plots.
-
Chronic Disease and Physical Inactivity Epidemiology Model Explorer (Python Dash dashboard)
- Built an interactive Dash application using the CDC BRFSS dataset to analyze how social determinants of health influence physical inactivity in chronic-condition populations.
- Deployed the app on Google Cloud Run with cohort building by chronic condition, BMI, sex, and income, plus model training and evaluation workflows.
- Integrated Gemini AI to generate model narratives, identify key risk drivers, and surface public health recommendations.
-
Analysis report: SDOH and Physical activity amongst people with diabetes
-
Predicting Glioma Grade from Gene Mutation Profiles (Python Streamlit Dashboard)
-
Tableau Public