SafeGraph AI: Explainable Disaster Preparedness Agent

SafeGraph AI

Inspiration

Disasters can cause severe damage, but the impact can often be reduced when people have the right knowledge and preparedness before and during an emergency. We were inspired by the need for a simple, accessible platform that does more than provide static disaster information.

We wanted to create a system where users can learn about disasters, test their knowledge, understand risks, receive alerts, and ask an AI assistant for explainable guidance in one place.

This led us to build SafeGraph AI: A Knowledge Graph and Agentic AI Framework for Explainable Disaster Preparedness and Emergency Response Education.

What We Built

SafeGraph AI is an AI-powered disaster preparedness and emergency response education platform designed as a phone-first experience.

The platform brings together three core technologies:

Knowledge Graph — represents relationships between disasters, causes, warning signs, effects, precautions, and emergency actions. Risk Analysis — provides disaster risk/status information that can be used to support timely awareness and alerts. Agentic AI — understands user questions, retrieves relevant knowledge, reasons over the available information, and produces explainable responses.

Instead of functioning as a generic AI chatbot, SafeGraph AI connects AI reasoning with structured disaster knowledge.

How It Works

The overall workflow is:

User ↓ User Query / Learning Activity ↓ Agentic AI ↓ Knowledge Retrieval ↓ Knowledge Graph ↓ Reasoning ↓ Explainable Response ↓ User Guidance

For example, when a user asks:

"What should I do during an earthquake?"

the system identifies the user's intent, retrieves relevant earthquake preparedness information from the Knowledge Graph, uses the available context for reasoning, and generates an understandable response.

This helps make the AI response more relevant, structured, and explainable.

User Experience

Users can access:

📚 Learn / Courses

Users can browse disaster-wise educational content covering:

• Introduction
• Causes
• Warning signs
• Effects
• Prevention
• Emergency preparedness
• Before, during, and after disaster guidance
• Emergency kits
• Supporting learning materials

📝 Quizzes

Users can select a disaster, take a quiz, receive a score, review correct and incorrect answers, and view their previous quiz attempts.

🤖 Agentic AI

Users can ask natural-language questions such as:

• "How should I prepare for a cyclone?"
• "What should I do during an earthquake?"
• "What items should I keep in an emergency kit?"
• "Why is this disaster warning classified as high risk?"

The Agentic AI uses the available structured knowledge to provide guidance and explanations.

🔔 Notifications

Users can receive relevant emergency alerts, disaster risk warnings, learning reminders, quiz results, and important system notifications.

📊 Progress Tracking

Users can monitor their learning progress and quiz performance from their dashboard.

Admin Dashboard

SafeGraph AI also includes an Admin dashboard that acts as an overview and analytics center.

Administrators can:

• Manage users
• Create and manage educational courses
• Add and manage quiz questions
• Categorize content by disaster
• Publish or unpublish educational content
• View quiz reports
• Monitor user activity
• Analyze course performance
• Analyze disaster-wise quiz performance
• Monitor AI query activity

This allows administrators to continuously maintain the educational content and understand how users are learning.

What Makes It Different

SafeGraph AI is not simply a disaster information website or a generic AI chatbot.

Its key idea is the combination of:

Knowledge Graph + Risk Analysis + Agentic AI + Smart Education

The Knowledge Graph provides structured relationships between disaster concepts. Risk analysis supports situation awareness, while Agentic AI uses the available knowledge to understand, reason, and explain information to the user.

This creates a complete learning and preparedness journey:

Learn ↓ Practice ↓ Ask AI ↓ Understand Risk ↓ Receive Guidance ↓ Prepare

How We Built It

Our project uses a modern application architecture with a frontend, backend APIs, database/knowledge layer, AI services, and deployment infrastructure.

The main technologies used in our implementation include:

Next.js — frontend application FastAPI — backend API services Neo4j — Knowledge Graph LLM / Agentic AI — natural-language understanding, retrieval, reasoning, and response generation PostgreSQL — application data Git / GitHub — version control and collaboration Vercel — frontend deployment and hosting

Learning

Through this project, we learned how to combine conventional application development with AI-based systems.

The major areas we worked on include:

• Designing a Knowledge Graph for domain-specific information
• Connecting structured knowledge with an LLM
• Designing an Agentic AI workflow
• Building retrieval and reasoning flows
• Developing disaster-focused educational content
• Designing dashboards and analytics
• Integrating frontend and backend services
• Deploying the application using cloud infrastructure
• Designing a phone-first user experience

Challenges We Faced

One of the major challenges was determining how to make AI responses useful and grounded in disaster-specific information instead of relying only on generic LLM responses.

We addressed this by introducing a structured Knowledge Graph and designing the AI workflow around knowledge retrieval and reasoning.

Another challenge was bringing multiple components together into one coherent platform. Courses, quizzes, notifications, risk information, AI assistance, user progress, and administration all needed to work as part of a single system.

We also focused on making the interface simple enough for users to quickly understand information during situations where clarity is important.

Future Scope

Our future improvements include:

• Stronger real-time disaster data integration
• More advanced risk analysis
• Expanded disaster coverage
• Improved AI grounding and explainability
• More personalized learning recommendations
• Enhanced mobile and on-device AI capabilities
• Offline access to critical preparedness content
• Additional multilingual support

Conclusion

SafeGraph AI aims to make disaster preparedness more accessible, interactive, and explainable.

By combining Knowledge Graphs, risk analysis, Agentic AI, courses, quizzes, notifications, and analytics, the platform transforms disaster preparedness from static information into an interactive learning and guidance experience.

SafeGraph AI — Learn. Understand. Prepare. Respond.

Built With

• agenticai
• artificialintelligenece
• disastermanagement
• fastapi
• generativeai
• knowledgegraph
• llm
• natural-language-processing
• neo4j
• next.js
• postgresql
• python
• smarteducation
• vercel