Building a benchmark for machine learning on type 1 diabetes data.
Gumbel-Based Active Sparse Mobile Crowd Sensing with Time Series Transformer
A learned Gumbel-noise sensor selection layer paired with a time series transformer, reducing reconstruction error on missing sensor data by up to 28%.
Ensemble Learning with Early Fusion of Kernel-Transformed and Classical Electrocardiogram Features for Chagas Disease Detection
Ensemble framework over AutoGluon, ECG-FM, FFT, and wavelet features for 12-lead ECG classification. Placed 39th in the 2025 George B. Moody PhysioNet Challenge.
Patched Forecasting with Gumbel-Based Selector for Sparse Mobile Crowd Sensing
Poster companion to the full paper, presented at IPCCC 2025 in Austin, TX.
- In progress Type 1 diabetes benchmark
- Oct 2025 — Child care quality dataset
A new AI/ML benchmark dataset of child care quality ratings, scraped from the Georgia Department of Early Care and Learning. Currently building a causal discovery graph via FPM and PC, with a DiCE counterfactual framework to rank the key determinants of quality across 245 provider attributes.