AI4ED Research Initiative
In emergency care, every second matters and every decision counts.
AI4ED develops trustworthy, human-centered AI that turns complexity into clarity, helping clinicians make better decisions, improve workflow, and deliver better patient care.
AI4ED Research Initiative
AI4ED is an interdisciplinary research initiative building clinically grounded AI systems for emergency department chart summarization and workflow simulation. We combine emergency medicine, machine learning, informatics, and engineering to study how AI can improve care delivery and operational decision-making.
Explore AI4ED
Interactive Demo
Explore how aligned language-model agents coordinate triage, assessment, and disposition decisions in a simulated emergency department. The demo makes the project tangible by showing how research ideas translate into a usable experimental environment.
Recent Publications Introduces an agentic emergency department simulator that combines LLM-driven clinician and patient behavior with data-driven operational modeling for workflow experiments. Demonstrates how clinician-in-the-loop fine-tuning boosts the quality and utility of ED chart summaries. Evaluates compact models for ED decision-support tasks, highlighting trade-offs between accuracy, latency, and deployability.EDSim: An Agentic Simulator for Emergency Department Operations
Fine-tuned large language models guided by physician feedback can improve patient chart summarization for emergency departments
Small Language Models for Emergency Departments Decision Support: A Benchmark Study
Project Roadmap
A high-level view of how capstone work, summer development, funding support, and publication goals connect across the AI4ED program. 
Funding Support
This work is supported by the Schulich Momentum Fund at the Schulich School of Engineering, University of Calgary. The initiative strengthens research impact, collaboration, student success, and innovation across engineering, and we gratefully acknowledge that support in advancing the AI4ED program.