Current direction

I build clinical analytics that connect patient data, model behavior, and decision-ready evidence.

EHR prediction Clinical NLP Model evaluation

Focus Areas

Prediction that can be checked

Models for emergency care, utilization, and patient trajectories, with attention to calibration, subgroups, and practical interpretation.

Clinical text as signal

NLP and multimodal workflows that combine structured variables with notes, narratives, and workflow context.

Evaluation before excitement

Clearer thinking about what models can support, where they fail, and how evidence should be reported.

Reproducible analytics

Analysis pipelines that make cohorts, variables, models, and reports easier to audit and revise.

Representative Work

Hospital admission prediction using gradient boosting and GPT-2DIGITAL HEALTH, 2025. DOI
Multimodal prediction of IV fluid utilizationPeerJ Computer Science, 2025. DOI · Code
Personalized prediction of EKG use in emergency careJournal of Personalized Medicine, 2025. DOI · Code

Coming Next

Short notes on clinical model evaluation, clinical NLP, EHR prediction pipelines, missingness, and real-world evidence workflows.