Disaster Preparedness & Research Lab · AI
MultiAgentCrisis — Multi-Agent Simulation & Policy Sandbox
10,000+
Agents Simulated
60 FPS
Tick Rate
98.5%
Log Compression
< 5s
AI Latency
The challenge
Disaster management planners relied on static models that failed to capture human psychological variables (panic, empathy, greed), cascading infrastructure failures, or real-time emergency crowd dynamics.
The engineering solution
We engineered MultiAgentCrisis — a large-scale computational simulation using Multi-Agent Reinforcement Learning (MARL) and a 3-tier Causal XAI Debriefing engine powered by DeepSeek AI. Simulates 10,000+ autonomous agents across GIS maps with policy sandbox testing.
Tech stack
Next.jsPythonFastAPIDeepSeek APINetworkXDuckDBMapbox
Want to build a system like MultiAgentCrisis?
Receive a transparent, itemized proposal within 48 hours.
Start a conversation →
