SafePath AI

An autonomous disaster-response agent that transforms emergency alerts into personalized action plans, evacuation guidance, shelter discovery, and family coordination.

Comment

SafePath AI

Inspiration

Every year, millions of people face floods, cyclones, heatwaves, and other natural disasters. While emergency alerts are often available, many people still struggle to understand what actions they should take, where they should go, and how to keep their families informed.

We were inspired by a simple question: what if an AI agent could do more than warn people? What if it could actively help them prepare, plan, and respond during emergencies?

SafePath AI was created to transform disaster information into clear, personalized action plans that help people stay safe when every minute matters.

What It Does

SafePath AI is an AI-powered disaster response agent that helps individuals prepare for and respond to emergencies.

The agent can:

  • Monitor disaster alerts and emergency information
  • Analyze risk levels for specific locations
  • Generate personalized evacuation plans
  • Recommend emergency supply checklists
  • Locate nearby shelters and safe zones
  • Maintain emergency contact information
  • Provide step-by-step guidance during critical situations

Rather than simply displaying warnings, SafePath AI acts as an intelligent assistant that helps users make informed decisions quickly.

How We Built It

SafePath AI was built using Google Cloud Agent Builder and Gemini to create an intelligent multi-step agent capable of reasoning through emergency situations.

The system uses MongoDB through MCP integration to store and manage:

  • User profiles
  • Emergency contacts
  • Shelter information
  • Disaster preparedness records
  • Alert history

Gemini analyzes incoming emergency information, evaluates potential risks, and generates personalized recommendations. The agent then coordinates multiple tasks, including evacuation planning, preparedness guidance, and emergency communication workflows.

Challenges We Ran Into

One of our biggest challenges was designing an agent that goes beyond a traditional chatbot. We wanted the system to actively assist users instead of simply answering questions.

Another challenge was structuring emergency-response workflows that remain easy to understand during stressful situations. We focused on creating clear, actionable recommendations that users can follow quickly.

We also worked on organizing location, shelter, and emergency-contact data in a way that supports fast decision-making.

Accomplishments That We're Proud Of

We are proud of building a project that focuses on real-world impact and public safety.

SafePath AI demonstrates how modern AI agents can move beyond information retrieval and support meaningful decision-making during emergencies. The project combines reasoning, planning, data management, and user-centered design into a single system.

Most importantly, we created a solution that has the potential to help people prepare for situations where timely guidance can make a significant difference.

What We Learned

Through this project, we learned how to design agentic workflows that perform multiple coordinated tasks rather than simple question answering.

We gained experience working with Google Cloud Agent Builder, Gemini reasoning capabilities, MCP integrations, and structured data management. We also learned the importance of designing AI systems that remain reliable and understandable during high-pressure situations.

What's Next for SafePath AI

Future versions of SafePath AI could include:

  • Real-time disaster data integrations
  • Voice-based emergency assistance
  • Multilingual support
  • Community-level preparedness tools
  • Automated family notification systems
  • Accessibility features for elderly and disabled users
  • Mobile deployment for broader accessibility

Our vision is to develop SafePath AI into a comprehensive emergency preparedness and response platform that helps communities become more resilient in the face of disasters.

Built With

Share this project:

Updates