Some ongoing research

A simulation-based inference workflow in which a working model learns from simulations and reverses the mapping for observed data.

Simulation-based inference for epidemic models

Likelihood-free inference and computational methods for complex mechanistic epidemic models.

Patient data remain private at healthcare sites while aggregate summaries are combined into shared evidence.

Hierarchical framework for federated data analysis

Bayesian hierarchical methods for evidence synthesis across federated data networks.

Conformal calibration uses a narrow adaptive interval to cover patient-level causal estimates at exactly matched coordinates.

Conformal prediction for individualized causal inference

Valid uncertainty quantification for heterogeneous causal effect learning.

CGM data feed time-series learning and mechanistic glucose-insulin dynamics in a hybrid forecasting model.

Hybrid ML + math model glucose prediction

Combining ML with mathematical models for glucose forecasting using continuous glucose monitor (CGM) data.