TriageAI: Multimodal AI for Emergency Dispatch

The Problem: Critical Decisions in Seconds

Every year, over 240 million 911 calls are placed across 5,748 Public Safety Answering Points (PSAPs) in the United States. Dispatchers are tasked with making life-or-death decisions in a matter of seconds, often with minimal information, panicked callers, and no automated support.

This high-stress environment leads to critical gaps. Studies show that a significant portion of high-severity events are misclassified during the initial call—for example, strokes mistakenly coded as simple "falls," leading to the dispatch of Basic Life Support (BLS) instead of Advanced Life Support (ALS). Missed treatment windows cost lives. Current Emergency Medical Dispatch (EMD) systems rely on rigid flowcharts or prohibitively expensive, closed-source enterprise software, leaving many suburban and rural dispatch centers without the modern, data-driven tools they need.

The Solution: A Multimodal "Second Opinion"

TriageAI is an open-source, multimodal machine learning pipeline designed to assess the severity of 911 calls in real-time. By analyzing both what is being said and how it is being said, TriageAI acts as an intelligent "second opinion" to validate and support a dispatcher's judgment.

Radical Transparency

Dispatchers need to know why an AI made a prediction. Our tool highlights the exact phrases and vocal cues that triggered a high-severity alert.