Clinical AI product leader and practicing emergency physician working at the research-to-production boundary for LLM-based healthcare AI. Co-founded Sayvant and leads documentation and evaluation work used across 100+ care sites and more than 1.1M charts. Builds physician-written rubrics, chart-grounded QA, dangerous-fabrication review, PHI-aware routing, HIPAA/BAA-aware deployment, and post-launch monitoring. Stanford MCiM graduate, Army veteran, FDA-regulated Class I device founder, and first author on 2026 medRxiv preprints spanning production documentation QA and video-linked airway procedural telemetry.
Product competencies: Product strategy and roadmap | LLM/model evaluation | Physician-written rubric design | Clinician-in-the-loop QA | Privacy and safety controls | HIPAA/BAA-aware workflows | Discovery-to-launch execution
Clinical AI Product Lead (Co-Founder)
Oct 2023 - Present
Owns clinical AI product strategy, requirements, LLM/model evaluation, physician-review workflows, and rubric development for documentation systems used across 100+ care sites and more than 1.1M charts. Built HIPAA/BAA-aware deployment support, PHI-aware routing, chart grounding, hallucination and dangerous-fabrication review, billing completeness, and model-output triage. First author on a production QA preprint across 13 hospital sites: 42 tracked complaints, zero regressions across 1,089 iterations, 282 binary checks, 166 deterministic repair rules, and 6/6 fabrication-class failures eliminated in ablation.
Founder & Product Lead
Oct 2018 - Present
Leads product strategy for a $95 single-use USB-C video laryngoscope platform while transitioning CEO responsibilities to incoming leadership. Owned discovery, requirements, patented design, FDA-regulated Class I execution, production hardware, pricing, commercial launch, clinical education, and field adoption across 400+ EMS agencies in 41 states, with $250K+ recorded sales revenue and 2,884 cumulative devices in field. Published a June 2026 medRxiv preprint and public evidence package for 30-run manikin computer-vision airway telemetry. Built operator systems for legal, shipping, inventory, pilots, AR, GTM, pediatric and bronchoscope roadmap expansion, and human-in-the-loop procedural guidance using connected video, telemetry, and trainer-confirmed safety fields.
Emergency Medicine Physician
Jan 2023 - Present
Kaiser Permanente
East Bay, CA
Practices emergency medicine in an integrated health system, keeping AI and airway product decisions tied to department constraints, clinician trust, and real patient care.
Stanford University School of Medicine: Master of Science, Clinical Informatics Management (Jun 2025 - Jun 2026, Stanford, CA) | University of South Florida College of Medicine: Doctor of Medicine (Aug 2013 - May 2017, Tampa, FL) | Eastern Kentucky University: Bachelor of Science, Biomedical Science (May 2009 - Dec 2012, Richmond, KY)
First-author preprint with Serge Klement and Ben Fedeles. The work audits a public post-reconciliation evidence package from 30 simulated airway runs using an IntuBlade device and iPhone workflow. The claim is intentionally bounded: video-linked structured JSON, QC status, app/model metadata, package hashes, and assigned-condition consistency for formative airway telemetry, not autonomous guidance, model accuracy, training effectiveness, clinical efficacy, or patient outcomes. DOI: 10.64898/2026.06.26.26356677.
First-author preprint with Justin Wiley and Mark Heslin. The work turns physician complaints from production AI documentation across 13 hospital sites into reproducible QA artifacts, driving 1,089 optimization iterations with zero regressions on resolved complaints. Evidence base: 42 tracked complaints, 282 binary checks, 166 deterministic repair rules, and 6/6 fabrication-class failures eliminated in ablation. DOI: 10.64898/2026.05.27.26353977v1.
AIMI poster Jun 3, 2026; AMIA oral presentation Nov 11, 2026
Stanford AIMI / AMIA 2026 / manuscript in preparation
Stanford MCiM capstone work with a public control repo, AIMI poster, and an AMIA 2026 oral presentation (session S99, abstract #15227) with the two-page abstract in the Annual Symposium Proceedings. The project pairs ambient patient signal with physician-authored decision traces so decision-support models are not trained only on downstream chart text. Manuscript and validation plan in preparation with Ashley Griffin and Mark Musen.
ICD-10-CM Coding with Retrieval and LLM Reranking
Accepted Sep 2026; poster, New York, Oct 2026
IEEE Healthcom 2026, MIHA workshop (AI and LLMs for Medical Imaging and Healthcare Applications)
Senior author with Geanderson Esteves dos Santos (first author), Justin Mardjuki, and Justin Wiley. Diagnosis-to-ICD-10-CM code selection that pairs frequency-aware retrieval over California HCAI code-frequency data with LLM reranking, plus synthetic emergency and hospital medicine examples and error analysis.
Health-System Implementation of AI in Acute Care Delivery
Presented May 5, 2026
HITLAB Spring Summit, Sayvant session
Sayvant session on acute-care AI rollouts, workflow ownership, clinician review, rollout friction, feedback loops, and product lessons from health-system deployment. Final title: Planning for the Future: Lessons Learned Partnering with Health Systems to Implement AI in Acute Care Delivery.
Sayvant research paper, methodology demonstration
With Justin Wiley. A transcript-versus-output diff benchmark in three layers (text only, Whisper end to end, Deepgram end to end), with HPI and plan fabrication partitions and bootstrap confidence intervals, run on the open primock57 UK primary-care corpus as a worked example. Code and checks are public as ScribeBench (github.com/napiermd/scribe-bench).
Sayvant research paper, in draft
With Ashley Griffin and Mark Musen. Pairs the patient presentation with the physician decision trace for 360 ED chest-pain cases, trains ACS, troponin, disposition and differential models, and reports a three-physician reviewer evaluation, including where the model over-calls ACS risk. Poster and live demo are linked from the paper page.
Introduces the Sayvant Quality Score (SQS): section-level LLM microjudges aggregated into HPI and MDM judges, validated against board-certified emergency physician scoring on 500+ de-identified charts and deployed across 70+ emergency and urgent care sites.
Research Focus & Collaboration Areas
Research outreach is focused on collaborators evaluating clinical AI in real care: ambient documentation QA, physician-grounded labels, acute-care decision support, clinical coding, implementation science, and human-in-the-loop procedural guidance. Best-fit projects combine governed clinical signal, clinician adjudication, careful validation, and publishable methods.
Research Interests
- Clinical AI evaluation and implementation science in emergency and hospital medicine
- Ambient documentation QA, dangerous-fabrication detection, and physician oversight
- Clinical conversation signals for risk stratification, decision support, and validation
- Physician-authored rubrics, labels, and adjudication workflows for model evaluation
- Clinical coding, structured labels, retrieval-plus-selector methods, and error analysis
- Video-linked procedural telemetry and human-in-the-loop guidance for airway and other high-risk procedures
Technical Capabilities
- Healthcare AI product management: strategy, roadmap decisions, problem discovery, requirements writing, launch planning, adoption, and post-launch quality
- Clinical AI study design: expected utility, prospective validation planning, implementation evaluation, and post-launch monitoring
- Physician-review workflow design: chart grounding, documentation reliability, hallucination risk, dangerous-fabrication review, and defensibility checks
- Clinical NLP and ambient documentation systems: conversation-to-chart workflows, extraction patterns, structured chart generation, and QA loops
- High-acuity workflow design across emergency medicine, hospital medicine, airway management, and procedural training
- Healthcare classification and retrieval systems: ICD-10 pipelines, frequency-aware retrieval, constrained selection, synthetic examples, and error analysis
- FDA-regulated device development, clinical workflow design, commercialization, and post-launch adoption
Clinical & Military Experience
Vice Chair, Assistant Medical Director
Dec 2021 - Sep 2023
Emanuel Medical Center
Turlock, CA
Led quality, throughput, documentation, patient experience, and sepsis improvement work for a 70,000-visit emergency department.
Emergency Medicine Physician
Sep 2020 - Dec 2021
Regional Medical Center
San Jose, CA
Practiced emergency medicine at a Level 2 Trauma Center with annual volume exceeding 100,000 visits.
Emergency Medicine Resident Physician
Jul 2017 - Jul 2020
Kendall Regional Medical Center
Miami, FL
Completed emergency medicine residency at a Level 1 adult and pediatric trauma center.
Sergeant (E-5), Combat Medic Specialist
Jun 2006 - Jun 2012
United States Army National Guard
Ashland, KY / Afghanistan
Led battalion-level medical operations, supervised medics, and provided combat casualty care alongside a forward surgical team in Afghanistan. Treated more than 300 documented casualties and received the Purple Heart, Combat Medical Badge, and Combat Action Badge.