GroundZero

Inspiration 🚀

In disaster scenarios like earthquakes, first responders often lack real-time visibility into what’s actually happening on the ground. Decisions about where to go and who to help first are made under extreme uncertainty. We wanted to build a system that turns chaos into clarity by giving responders a live, intelligent view of the situation.

What it does 🧠

GroundZero is an AI-powered 3D disaster intelligence platform that simulates and visualizes earthquake impact in real time. It displays an interactive map where buildings are color-coded by triage severity based on estimated damage, occupancy, and risk. Users can click into buildings to see detailed breakdowns and recommended response priorities. The platform also includes a first-person navigation mode and uses multiple AI agents to predict damage, simulate how impact spreads, and guide optimal response routes.

How we built it 🛠️

We built GroundZero using React and Next.js for the frontend, with Mapbox powering the 3D map and visualization layers. On the backend, we used Python with FastAPI to run a multi-agent system responsible for damage prediction, risk propagation, and route optimization. The system is powered by a graph-based model where buildings, roads, and infrastructure are connected, allowing agents to update and reason over a shared state in real time.

To make our simulations more grounded and data-aware, we integrated OpenClaw to dynamically fetch and enrich building-level context. OpenClaw allows us to pull structured and unstructured data about selected locations, such as building metadata, surrounding infrastructure, and contextual risk signals. This data feeds directly into our agent system, improving triage scoring and enabling more realistic, context-aware decision making when responders interact with specific buildings.

Challenges we ran into ⚠️

One of the biggest challenges was simulating realistic disaster impact without access to real-world data. We had to design our own assumptions and models for damage and risk. Rendering large numbers of buildings in 3D without performance issues was another challenge. We also had to rethink our architecture from agents simply “talking” to each other to instead updating a shared graph. Balancing complex data with a clean and intuitive UI was also difficult.

Accomplishments that we're proud of 🏆

We’re proud of building a fully interactive 3D disaster map that feels like a real command center. Our multi-agent system doesn’t just generate outputs, it collaborates through a shared graph to simulate real-world dynamics. We also created a system that explains why decisions are made, not just what to do. Bringing together AI, mapping, real-time systems, and external data enrichment through OpenClaw into one cohesive product was a big achievement.

What we learned 📚

We learned that visualizing AI decisions is just as important as making them. Multi-agent systems become much more powerful when they operate on shared data rather than conversations. We also saw how important performance and UX are when dealing with complex systems. Most importantly, we learned how to design technology around real-world decision-making.

What's next for GroundZero 🔮

Next, we want to integrate real-time data sources like USGS and weather APIs to improve accuracy. We plan to expand beyond earthquakes to other disasters like wildfires and floods. We also want to add predictive modeling for secondary effects and enable multi-user collaboration for coordinated response teams. Ultimately, we aim to scale GroundZero into a full disaster response platform that can be used for both real-world deployment and training simulations.

Built With

• chatgpt
• claude
• google
• maps