Work from the
2026 cohort.

SHARPEN provides virtual standardized patients for students to practise clinical conversations, with real-time interaction, immediate feedback and a range of cases.

The team demonstrated the platform as a way to make clinical practice more accessible. GI-Consent and VSP-QAT, listed separately below, examine how students learn with the platform and how the virtual-patient experience should be assessed.

Grand Demo Day 2026, slides 13–19, 32

Can practice with an AI virtual patient help students learn to deliver informed consent? GI-Consent compared AI practice with peer role play.

Eleven students practised with AI virtual patients and eleven with peers. The presentation reported no statistically significant differences between groups in consent performance (p = 0.27), communication (p > 0.05) or confidence (p = 0.80). These results describe a small comparison and do not establish equivalence.

Grand Demo Day 2026, slides 20–22

A virtual patient needs to do more than hold a conversation. VSP-QAT evaluates clinical accuracy, realism, communication, consistency and educational value.

The team built a quality-assessment tool and outlined a blinded evaluation of recordings from deliberately high- and low-quality virtual patients. The proposed design includes 12 student volunteers, four recordings and ten blinded reviewers. The next step described in the deck is a pilot study.

Grand Demo Day 2026, slides 23–31

Tipoca reviews a recording of a clinical procedure against a standard operating procedure. It uses a vision-language model to assess individual steps and helps instructors review a student’s performance.

The team presented a working prototype with timestamped annotations and an editable report. Their development work included the evaluation pipeline, application architecture and interface. Testing highlighted variability in model responses. The stated next steps were a maintenance and deployment pipeline and production deployment.

Grand Demo Day 2026, slides 33–45

This project tests whether vision-language models can understand clinical procedures and assess performance from video, beyond recognizing a static medical image.

The presentation uses NurViD videos to evaluate procedure, action-step and joint classification, and AIxSuture videos to assess suturing proficiency and rubric scores. Results show different performance across models and tasks, with examples of commonly confused procedures and actions. The work examines what these models can reliably contribute to clinical education.

Grand Demo Day 2026, slides 46–58

Sharpen Box records a clinical encounter for an AI-supported review. Instructors can review suggested feedback and approve the points a student should work on next.

The demonstration connects a recording device, a video-transfer application and a performance-review interface. Its aims are timely personalized feedback, a way to follow improvement over time and less administrative work for instructors. A separate PAS validation study examines assessment accuracy and the experience of learners and institutions.

Grand Demo Day 2026, slides 59–66

The PAS study asks whether an AI assessor can match a human examiner and support continuing formative assessment in medical education.

The presentation sets out objectives to validate accuracy, measure learner growth and assess institutional value. It names clinical-performance and communication measures alongside technology-acceptance instruments, and outlines a comparison design. Integration within SHARPEN and a wider assessment network are described as future directions.

Grand Demo Day 2026, slides 68–75

EMBER examines burnout among Canadian medical-school applicants and asks how psychological resilience may influence their experience of the application process.

The survey covers demographics, academics, extracurricular activities and resilience. The recruitment plan includes pre-medicine societies, university groups and social media. The presentation describes survey development, recruitment and planned analysis, rather than completed national findings.

Grand Demo Day 2026, slides 78–96

Why Learn asks how AI literacy relates to students’ readiness to use AI in healthcare and innovation.

The study targets 200–300 learners and students. Survey topics include AI knowledge, perceptions of healthcare uses, ethical concerns, education, career motivation and AI-assisted workflows. Its expected association between self-perceived knowledge and readiness is a hypothesis; the deck does not present completed survey findings.

Grand Demo Day 2026, slides 97–105

The review asks whether real-time computer-aided detection improves trainees’ adenoma detection, and whether any benefit persists once AI assistance is removed.

The planned search spans ten sources from 2017 onward. Two reviewers screen the literature and extract trainee characteristics, AI systems, detection outcomes and performance without AI. The intended analysis separates assisted performance from learning transfer. No pooled result was presented in these slides.

Grand Demo Day 2026, slides 106–117

This study asks whether AI assistance changes a learner’s ability to retain a newly learned clinical skill. It uses arterial blood gas interpretation as the task.

Novices first complete standardized training and reach a minimum performance standard. The proposed trial then assigns them to no AI, unrestricted AI or Socratic AI that guides their reasoning without giving diagnostic answers. The primary endpoint is performance 24 hours later with AI removed. Cognitive load, reasoning quality, confidence and AI trust are secondary outcomes.

Grand Demo Day 2026, slides 118–130

Mantaculus asks learners to answer a question, rate their confidence and explain their reasoning. Adaptive feedback and follow-up questions respond to gaps in understanding.

The proposed randomized study compares Mantaculus with traditional learning among adult pre-medicine undergraduates. It includes a pre-test, a post-test and a delayed post-test to assess knowledge retention. The research also asks whether learners become more accurate in judging what they know. The presentation did not report trial outcomes.

Grand Demo Day 2026, slides 131–142

The IPCT project examines primary-care attachment for patients with medical and social complexity. It asks who should receive referrals, whether attachment changes subsequent healthcare use and whether the model delivers long-term economic value.

The team outlined a data-based definition of complexity, a 1:3 matched-cohort comparison and a health-economic model. Outcomes include emergency-department revisits, inpatient admissions and mortality. Preliminary service figures in the deck provide context for the proposed evaluation; they are not a completed causal estimate of the program’s effect.

Grand Demo Day 2026, slides 145–162

GEO-EoE proposes a map of regional eosinophilic esophagitis incidence in Canada, with a model to estimate incidence in places where data are missing.

The work investigates possible associations with environmental and geographic factors, including air and water quality, allergens and proximity to gastrointestinal clinics. The presentation outlines data collection, tool development, a retrospective risk-factor study and predictive modelling. These are research aims, not established associations.

Grand Demo Day 2026, slides 163–176

This project asks whether HealthBench reference answers agree with clinical practice guidelines. It focuses on gastroenterology and hepatology.

The team’s approach is to assemble a guideline corpus, identify relevant HealthBench conversations, match them to recommendations and score concordance. Proposed future work includes a new benchmark and expansion to other specialties. The presentation does not report a completed concordance analysis.

Grand Demo Day 2026, slides 177–195

ContextCare uses AI to turn complex discharge information into a patient-friendly summary before a person leaves hospital. The project pairs the software with a study of discharge communication.

The demonstrated workflow includes health-literacy assessment, simplified summaries and a teach-back assessment. Additional features shown include multiple languages, audio summaries and a review of identifying information before simplification. The proposed study evaluates feasibility and patients’ comprehension and retention; at-home access and clinical integration are future directions.

Grand Demo Day 2026, slides 196–214