Universal CPR AI Assistance Response Engine (U-CARE) Development and Evaluation

Overview

Each year, over 350,000 people in the United States experience out-of-hospital cardiac arrest, and nearly 90% do not survive. Early cardiopulmonary resuscitation (CPR) can triple survival rates. Unfortunately, bystander CPR is correctly performed in fewer than 40% of cases, with three major problems persisting: lack of widespread training, decision paralysis, and recall bias in moments of emergency. U-CARE is an integrated, low-cost smartphone-based system that serves as a universal CPR guide in real-life emergency and training environments, and was evaluated using a novel, comprehensive multimodal evaluation framework. U-CARE is inspired from American Heart Association grant that I had received and comprises of an adaptive large language model (LLM)-powered verbal guidance system that provides objective next best steps and real-time feedback in multiple languages, coupled with a computer vision model that continuously assesses CPR technique and on compression rate, depth, and hand placement. By providing direct guidance in real-time scenarios, U-CARE mitigates the major problems that persist in CPR administration. However, the evaluation of such multimodal tools in this and other disciplines has remained a challenge, as traditional evaluation criteria of BERT and others do not offer insight into the quality of model responses, so a novel framework, HumanELY, was developed to quantify performance across disciplines on topics of Relevance, Coverage, Coherence, Comparison, and Harm. HumanELY, along with MedHELM from Stanford University (a benchmark for evaluating LLMs on real-world clinical tasks with a taxonomy, validated by 29 clinicians from 15 specialties), serves as a combined approach to evaluating U-CARE multimodal systems that enable universal access to CPR training and real-time guidance for cardiac arrest, saving countless lives.

Competition history

• CSEF 2026 Behavioral & Social Sciences (Senior Division) · Entry S-03-33

Related projects

Closest projects by meaning, across every fair and year in the corpus.

Browse more like this

• All Behavioural and Social Sciences projects
• CSEF Behavioural and Social Sciences projects
• Every CSEF 2026 project