Back to Projects
sosAI – Real-Time AI-Powered Emergency Call Assistant
AI Assistant
Emergency Tech
Real-Time AI
Voice Recognition
Hackathon
Natural Language Processing
Translation
Shipfast x Unaite x 42AI Hackathon (Paris, France) • July 2025
AI Engineer
Project Overview
Built at the Shipfast x Unaite x 42AI Hackathon in Paris (July 2025), sosAI is an AI-powered assistant for emergency call centers. The system transcribes calls in real-time, detects multiple languages, and supports dispatchers with instant AI-generated advice, protocol prompts, and smart question suggestions. The result: faster, more accurate emergency responses, even in complex, multilingual scenarios.
Challenges
- •Integrating real-time speech-to-text and multilingual detection with low latency
- •Designing a UI that surfaces critical AI-generated information without overwhelming the dispatcher
- •Generating dynamic, context-sensitive question prompts for improved emergency triage
- •Simulating realistic emergency call flows and system activity for demo/testing
- •Ensuring the system adapts to diverse emergency scenarios and multiple languages
Key Achievements
- •Built a production-ready real-time emergency call assistant in under 10 hours
- •Delivered seamless live transcription and translation detection with OpenAI Whisper
- •Implemented AI-generated emergency protocols and dispatcher suggestions via Claude API
- •Designed a modern, responsive frontend using Lovable (TypeScript code generation)
- •Demonstrated significant reduction in dispatcher decision time and increased situational awareness
Technologies Used
OpenAI Whisper
Anthropic Claude API
Lovable (Frontend Generation)
TypeScript
React