This research introduces a specialized multi-agent LLM system designed to provide context-specific information and decision support during extreme natural hazard events, particularly wildfires.

• Combines Retrieval-Augmented Generation (RAG) with a multi-agent architecture to overcome LLM limitations in specialized domains
• Creates domain-specific knowledge bases that enable more accurate and contextual responses than generic LLMs
• Demonstrates meaningful improvements in hazard assessment, evacuation planning, and resource allocation
• Provides a blueprint for AI systems that can enhance community resilience against natural disasters

Security Impact: This system addresses critical gaps in emergency management by delivering specialized knowledge during time-sensitive crisis scenarios, potentially saving lives and infrastructure.

A RAG-Based Multi-Agent LLM System for Natural Hazard Resilience and Adaptation