Clinical AI portfolio
Emergency physician and founder building clinical AI where failure is expensive: documentation, emergency care, airway management, and decision support.
My lane is clinical AI product work: define the clinical task, design the workflow, evaluate failure modes, and keep the system tied to the patient record.
The public record is strongest where it should be strongest: deployed documentation systems, acute-care clinical experience, patents, publications, and operator work.
100+
care sites using Sayvant documentation systems
800k+
structured charts generated across Sayvant workflows
70k+
annual ED visits in a department I helped lead
$95
single-use USB-C video laryngoscope platform
I lead clinical AI work for documentation systems used in emergency medicine and hospital medicine. The work is less about making notes sound polished and more about making them accurate, inspectable, and safe for physicians to sign.
My work sits between clinical judgment, product behavior, workflow, implementation, and evaluation. That is where clinical AI usually succeeds or fails.
Physician-review workflows for ambient documentation, including chart-grounding checks, failure analysis, billing completeness, and legal defensibility review.
// Sayvant clinical AI documentation work
// Stanford MCiM clinical informatics work
// ED chest pain analytics project
Extraction-first documentation systems that reduce unsupported note content, modality drift, and fabricated findings.
// Sayvant clinical AI pipeline
// Clinical NLP patterns repository
// Model migration and regression testing work
Emergency department leadership and founder/operator work that keeps AI grounded in who acts, when, with what liability and workload.
// Emergency medicine practice
// Vice Chair / Assistant Medical Director experience
// Founder roles at Sayvant and IntuBlade
Airway technology and procedural support built around cost, access, FDA reality, and acute-care deployment.
// IntuBlade
// Issued patents
// AI-enhanced depth perception and anatomical highlighting patent filing
Repositories and writing that show the clinical AI lane without exposing customer data.
Architecture patterns for reliable clinical documentation systems, extraction-first prompt design, and validation templates.
Stanford MCiM practicum on ED chest pain risk stratification from clinical conversations and structured clinical features.
Public writing on why clinical documentation systems must stay grounded in what was actually said and done.
Public writing on model migration, regression risk, and why clinical documentation needs evaluation beyond fluency.
The clinical AI leaders who matter will not be the people who can talk the longest about models in the abstract. They will be the people who can define the clinical task, evaluate failure modes, understand workflow, protect patients from false confidence, and still ship systems that make care better.