Getting Started
AI4ED studies how clinically grounded AI can reduce information friction and operational pressure in emergency care. The project combines large language models, workflow simulation, and close clinical collaboration to explore tools that are technically rigorous, operationally relevant, and realistic to evaluate.
The goal of this page is to orient collaborators, students, and technical contributors before they use the simulation repository or review the public demo.
Emergency care is high-pressure, fragmented, and information-heavy.
Emergency departments operate under sustained demand, limited time, and incomplete information. AI4ED was created to study how AI systems can reduce avoidable friction without losing clinical credibility or operational realism.
Overcrowding and demand
Patient volumes frequently exceed the capacity of emergency departments, creating persistent pressure on flow and decision-making.
Throughput inefficiency
Delays emerge across triage, assessment, documentation, handoffs, and disposition, making system improvement difficult to test safely.
Documentation burden
Physicians and nurses spend significant time reconstructing patient history from fragmented records and repetitive chart navigation.
Clinical information friction
Past medical history is often difficult to synthesize quickly in urgent care settings, even when the data technically exists.
AI4ED is organized around two connected workstreams.
Chart summarization for emergency medicine
We develop LLM-based summarization workflows that help emergency clinicians review complex patient histories more efficiently, with an emphasis on clinical usefulness, readability, and human oversight.
Agentic emergency department simulation
We build workflow simulators that make it possible to test interventions, study bottlenecks, and evaluate operational ideas before they are introduced into practice.
The project is intentionally cross-functional.
AI4ED brings together emergency medicine physicians, nurses, software engineers, machine learning researchers, data scientists, social scientists, and students. That mix helps the project stay technically strong, clinically grounded, and aligned with real-world implementation constraints.
Clinical direction
Emergency medicine collaborators help define what information matters, where workflow pressure occurs, and what a usable AI-supported process would need to respect.
Technical development
Engineering and machine learning contributors build simulation, summarization, evaluation, and repository infrastructure that can be inspected and improved over time.
Evaluation and translation
Health-systems and implementation perspectives keep the work tied to evidence, practical constraints, and the difference between a promising demo and a deployable tool.
Download or clone the AI4ED simulation repository.
The GitHub repository contains the simulation codebase, setup instructions, and demo guidance needed to initialize the AI4ED environment locally.