On-Premises AI Emergency Triage System

This on-premises AI emergency triage system is a smart triage-support tool built specifically for emergency departments, designed to help medical staff make more accurate triage-level decisions under high-pressure, high-volume conditions, based on consistent, evidence-backed rules. Built around a large language model, it combines Taiwan's current five-level emergency triage standard with vital-sign data, integrating a patient's chief complaint, symptom description, and physiological data to produce a reference-quality triage recommendation.

In practice, staff enter the patient's chief complaint, key symptoms, and initial vital signs; the system automatically matches them against critical-sign and grading criteria using its rules and trained model, proposes a triage level, and shows the reasoning behind it. This preserves medical staff's professional judgment as the final authority while providing a repeatable, traceable support tool that reduces variance caused by differences in experience. The system is deployed entirely on-premises — the compute core and model can run on an in-hospital server or a designated closed network segment — so patient data and clinical records never need to leave for an external cloud, meeting healthcare's strict requirements for privacy, security, and compliance.