UW Emergency Medicine Trauma Case Conference, ED Thoracotomy, Pediatric Hematuria, and Evaluating Predictive AI
Two trauma cases, plus a pragmatic approach to evaluating predictive AI models at the bedside.
Hello, World!
Medical AI is easy to prototype and hard to ship. I specialize in shipping it.
I’m a physician-engineer who takes medical AI from concept to production in real Health IT environments. I work across the full lifecycle, from identifying high-value clinical problems and developing models, to integrating them into production workflows (including Epic) with monitoring, safety guardrails, and measurable operational impact.
I’ve led end-to-end development and implementation of clinical models that have run millions of times across hundreds of thousands of patient encounters. I bring decade-plus experience across healthcare data, enterprise EHR workflows, and applied machine learning, plus the clinical context to build tools clinicians actually trust.
All models are wrong, but some are useful.
- George Box -
If you need someone who can understand the clinical problem, find the right EHR data, build the model, and implement it safely in production, let’s talk.
Recent Blog Posts
AI Infrastructure: Technical Integration Testing
How do we ensure what we built actually works in production?
AI Infrastructure Example: C. difficile Infection Risk
Discussion of the technical integration for an in-hospital infection risk stratification model.
Healthcare AI Infrastructure - To be deprecated
Infrastructure is king.
M-CURES: AI Infrastructure for Predicting COVID-19 Deterioration in Hospitals
A detailed exploration of the development and implementation of M-CURES, a COVID-19 in-hospital deterioration model, highlighting the integration of AI model...
Medicine is a science of uncertainty and an art of probability.
- William Osler -