ERA – Emergency Reasoning Agent

• 

  Severity 3

• 

  Severity 1

• 

  Severity 4

• 

  Severity 5

• 

  Severity 2

• 

  Home Screen

ERA – Emergency Reasoning Agent

Inspiration

Coming from India, where emergency incidents are frequent and early information is often incomplete, we wanted to build an AI system that can reason quickly from minimal inputs such as a short description or a single image.

What It Does

ERA analyzes emergencies using text, images, or both to classify the situation, estimate severity, highlight risks, and suggest immediate actions in a structured and explainable format.

How We Built It

ERA uses a FastAPI backend and a React frontend, with Google Gemini’s multimodal AI powering the reasoning engine. The system is designed to produce structured outputs rather than free-form responses for clarity and reliability.

Challenges & Learnings

We faced challenges handling incomplete inputs and calibrating severity without overestimation. We learned that multimodal AI works best with clear reasoning roles, structured outputs, and simple designs in high-stakes scenarios.

What’s Next

Future work includes improving severity calibration, adding multilingual support, integrating with real emergency workflows, and enhancing explainability while keeping humans in control.v

Built With

• fastapi
• gemini
• react