Research program
Clinical and biomedical models that state what the evidence supports.
I study patient-level reliability, causal structure, and biological constraints. The goal is not a higher metric alone. It is evidence that can be interpreted and used responsibly.
Overview
Three connected areas, one standard for evidence.
My work spans medical images, electronic health records, physiological data, and biomolecular measurements.
Across these settings, I use causal reasoning, calibrated uncertainty, mechanistic structure, and reproducible computation to test whether a model is useful for the decision at hand.
Trustworthy multimodal clinical AI
Find the patients and examinations where a model fails.
At MIT Critical Data, I evaluate diagnostic-imaging architectures and connect image-level model errors to high-acuity clinical context. The analysis focuses on individual examinations and patients, because cohort averages can hide concentrated harm.
- Calibration and uncertainty quantification
- Consensus-based harm scoring across model ensembles
- Subgroup, fairness, and instance-level error auditing
- Encounter-level linkage of MIMIC-CXR and MIMIC-IV
- Threshold and severity-sensitivity analysis
Causal inference & patient-specific decisions
Estimate relationships, then report how stable they are.
My causal-inference work combines ensemble Bayesian network learning, conditional-independence testing, MCMC exploration of graph structures, and bootstrap stability. BaMANI grew from my doctoral research on heterocellular networks in cancer.
The same discipline informs my work on epilepsy-surgery outcomes. Diagnostic performance is interpreted with subgroup patterns, structural missingness, sensitivity analyses, and clinically meaningful thresholds.
- Bayesian and ensemble causal-network inference
- Bootstrap edge stability and multiple-testing control
- Penalized regression and prespecified sensitivity analysis
- Failure analysis for patient selection and treatment counseling
Mechanistic & multiscale scientific ML
Use biological structure to constrain prediction.
At Duke, I connect agent-based and differential-equation models with machine learning. Gradient-boosted trees, Gaussian-process surrogates, and physics-informed neural networks make mechanistic models faster to explore while retaining uncertainty.
In the NIH-supported Duke–Weill Cornell congenital CMV project, these methods support maternal–fetal immune modeling, vaccine-efficacy analysis, and cross-species immune-cell alignment.
- Mechanistic simulators, ODEs, and agent-based models
- Gaussian-process and machine-learning surrogates
- Physics-informed neural networks
- Optimal transport with relaxed marginals
- Docker and SLURM/HPC reproducibility
Methodological backbone