An AI-powered assistant to support nurses during the emergency department (ED) triage process. The system will operate as a “co-pilot,” actively listening to the nurse-patient interview via a microphone. It will provide real-time suggestions for follow-up questions, highlight potential red-flag symptoms, and propose a preliminary triage classification based on established protocols (e.g., ESI, Manchester). The core design principle is to augment the nurse’s clinical judgment, not replace it, by reducing cognitive load and administrative tasks (like typing) while improving consistency and patient safety.
Client
The client is a Vue.js application built with Vite, providing the user interface for the AI-powered triage assistant. It includes components for chat conversations, audio transcription, triage summaries, and dashboard views.
Server
The server is a Python backend using Poetry for dependency management and FastAPI for the API. It integrates AI models for speech transcription (Vosk, Whisper), conversation analysis (Llama, OpenAI), and emergency severity assessment. It includes database management with SQLite and Alembic for migrations.
- Full Stack Developer