CrisisResponse

AI multi-agent system that guides people through life-threatening emergencies in real time step by step, in any language, anywhere in the world. Powered by GPT-5.6.

Comment

Inspiration

Every second counts in an emergency. People die not because help was unavailable but because they did not know what to do in the critical first minutes. We built CrisisResponse because we believe AI should serve humanity in its most vulnerable moments. Watching news about accidents and disasters where people had no guidance inspired me to build something that could actually save lives.

What it does

CrisisResponse is an intelligent multi-agent AI system that guides people through real life-threatening emergencies in real time step by step, in any language, anywhere in the world. Users describe their emergency, upload a photo if available, and instantly receive survival instructions, nearest emergency services, country specific contact numbers, and AI generated guidance all within seconds.

How we built it

We built CrisisResponse using 5 specialized AI agents powered by OpenAI Agents SDK and GPT-5.6. Each agent has one expert job:

-Situation Agent — understands emergency type and severity using GPT-5.6 and GPT-4o Vision for image analysis.

  • Location Agent — finds nearest hospitals and emergency services via OpenStreetMap Guide Agent generates medically accurate step by step survival instructions.
  • Contact Agent — retrieves emergency numbers for the user's country globally.
  • Update Agent — monitors and adapts guidance as the situation evolves.

Tech stack: OpenAI Agents SDK, GPT-5.6, GPT-4o Vision, FastAPI, Next.js, OpenStreetMap, deployed on Render.

Challenges we ran into

  • Orchestrating 5 agents that communicate cleanly and pass state correctly.
  • Integrating Computer Vision alongside text input without breaking the pipeline.
  • Making guidance genuinely medically accurate and not generic.
  • Handling multiple emergency types with one unified pipeline.
  • Deploying reliably under time pressure.

Accomplishments that we're proud of

  • Built and deployed a fully working multi-agent AI emergency platform.
  • Computer Vision integration analyzes photos for more accurate guidance.
  • Predictive agent warns about escalation before the situation worsens.
  • System works in any language globally.
  • Live, deployed, and accessible right now.

What we learned

  • How to architect multiple AI agents that work together seamlessly.
  • How to integrate Computer Vision into an agentic pipeline efficiently.
  • How to build and deploy a full stack AI application under time pressure.
  • That the most impactful AI serves people in their most vulnerable moments.

What's next for CrisisResponse

  • Automatic SMS alerts to family members.
  • Offline mode for poor connectivity areas.
  • Partnership with Red Cross and UNHCR.
  • Mobile app for iOS and Android.
  • Real time multi language voice guidance.

Built With

Share this project:

Updates