Funded by UT REAL Health AI · University of Texas System
The MAPLES AI grading platform is being prepared for multi-site validation across UT System medical schools, with engagement through the UT Health Intelligence Platform underway and funding from the UT-REAL Health AI initiative.
Partnering across the UT System to validate AI-enabled clinical assessment.
Milestones, publications, and announcements from the project.
Watch Minhan Park and Licheng Yi demonstrate Wayfinder beyond rubric authoring: upload an encounter with media, create a group, start a grading run, check status, and summarize results from chat.
Watch Minhan Park and Licheng Yi demonstrate Wayfinder: create a structured rubric from a short description, enhance it with accept/reject suggestions, transform it for a new scenario, and continue in chat.
On Healthcare AI Pioneers, Thomas Dalton and Andrew Jamieson discuss UT Southwestern’s AI-enabled OSCE assessment work: rubrics, multimodal evidence, staged rollout, human review, and precision education.
Collaborating across the UT System to validate AI-enabled clinical assessment.
View all participating sites →Our Mission
Our mission is to give medical educators better tools so they can give students better feedback — faster, more consistent, and at a scale that wasn't possible before.
Share strategies and best practices for designing AI-compatible OSCE rubrics, while allowing each school to customize assessment to their own educational philosophy and clinical needs.
Prepare MAPLES for governed multi-site deployment while evaluating UTSW-managed, site-specific, and UT Health Intelligence Platform (UT-HIP) infrastructure options. No cross-site production host has been selected.
Conduct rigorous validation studies comparing AI grading accuracy against expert human raters across diverse patient scenarios and institutions.
Define institution-specific hosting and transfer patterns, including a UT-HIP feasibility review, subject to site-specific IRB, data-use agreement, security, and hosting decisions.
Six UT medical schools sharing rubrics, best practices, and lessons learned — building a community of practice around AI-assisted clinical education.
Build on related NEJM AI and JMIR AI work, publish multi-site findings in peer-reviewed venues, and present project results at national conferences.
UTSW production, a preprint agreement result, and the six-school consortium.
From UTSW production to governed multi-school validation.
Phase 1: UTSW Proof-of-Concept (Complete)
AI grading system developed and deployed in production at UT Southwestern, with 7,000+ encounters graded. The grant proposal estimates that the six participating medical schools collectively represent more than 3,200 medical students. A 300-encounter retrospective medRxiv preprint reported three-camera AI grading agreement of quadratic weighted kappa = 0.830 against a physician-adjudicated reference, compared with 0.732 for standard human evaluators against the same reference.
Phase 2: Site Onboarding & Approval Routing (Current)
Funded by the UT REAL Health AI Pilot Program, March 2026 ($300K, 18 months). Award setup is complete. Partner sites are moving through local IRB review, agreement routing, data preparation, and infrastructure planning at different stages. UTSW and UT-HIP technical feasibility work is also underway.
Phase 3: Initial Multi-Site Validation (Upcoming)
Begin validation with the earliest-ready, authorized site data packages while continuing site-specific technical setup and governance review.
Phase 4: Broader Validation & Dissemination
Cross-institutional validation studies, publication of multi-site results, open-source governance playbook, and framework for national adoption.
Born at the UT Southwestern Simulation Center, the MAPLES platform has graded over 7,000 clinical encounters in production. A 300-encounter retrospective multimodal preprint reported three-camera AI grading agreement of quadratic weighted kappa = 0.830 against a physician-adjudicated reference, compared with 0.732 for standard human evaluators against the same reference. Related work is published in NEJM AI and JMIR AI. Now, through the UT-REAL initiative, we're preparing that capability for governed validation across additional UT medical schools.
