Research
Published research.
Preprints on machine learning for healthcare, cardiology, and credit, each with open code and data.
Scholar →Interpretable Machine Learning for Child Stunting and Wasting Risk: A Cross-National Evidence Base for Programme Monitoring and Targeting
An XGBoost classifier on 880 country-survey observations across 150 countries separates WHO high-burden stunting observations with an AUROC of 0.948, with SHAP finding WASH access, GDP per capita, and under-5 population as the dominant drivers, a result with direct implications for programme monitoring and targeting.
Explainable Credit Default Prediction for Microfinance: A Gradient Boosting and SHAP Study on Consumer Credit Data
Three classifiers on 150,000 borrowers, with SHAP showing that revolving utilisation and delinquency history carry most of the signal, and all tree models calibrating poorly, a result with direct lending implications.
Compact 1D Residual Networks for Efficient 12-Lead ECG Classification: A PTB-XL Benchmarking Study
A compact 1D ResNet at 1M parameters and 4 MB matches a full-scale model on PTB-XL diagnostic superclass classification, reaching a macro AUROC of 0.9253 at 8.65 times fewer parameters, small enough for edge deployment.
Self-Supervised Early Warning of Critical Cardiorespiratory Events in Neonatal Vital-Sign Telemetry: A Pilot Feasibility Study
Self-supervised prediction of imminent desaturation from continuous neonatal telemetry, where a GRU reaches an AUROC of 0.70 to 0.96 across held-out patients with a median warning lead time of 80 seconds.
Have something in mind?
Book a short call and tell us what you are building.