StrataWatch

StrataWatch is an AI-powered disruption intelligence platform that detects emerging global risks, predicts cascading infrastructure failures, and simulates their impact down to individual buildings.

Comment

Inspiration

Global crises rarely arrive without warning. Long before disasters unfold, signals begin to accumulate across the world: seismic activity increases, environmental hazards emerge, infrastructure systems show stress, and news sentiment shifts.

The problem is not a lack of data — it is the lack of synthesis.

Today, these signals exist across dozens of disconnected systems: earthquake feeds, hazard databases, weather alerts, and geopolitical monitoring platforms. Governments and specialized analysts have tools to interpret this information, but civilians, journalists, and even many emergency planners are left without a unified view of what is happening.

Meanwhile, modern crises rarely remain isolated. Power grids affect hospitals. Ports affect supply chains. Weather affects transportation. One disruption quickly propagates through interconnected systems.

We built StrataWatch to explore a simple idea: what if we could continuously monitor global signals, model how disruptions cascade across infrastructure networks, and simulate their impact all the way down to individual buildings?

Our goal was to connect global intelligence with local preparedness.

What it does

StrataWatch is an AI-powered disruption intelligence and emergency simulation platform that monitors emerging risks around the world and translates them into actionable situational awareness.

At the global level, the platform displays a live hex-based risk heatmap, continuously updating as new signals arrive. These signals are ingested from multiple sources including USGS earthquake feeds, NASA environmental hazard events, Open-Meteo weather disruptions, conflict and news signals, and simulated fallback signals for demo stability.

Each region receives a dynamic risk score derived from a probabilistic aggregation pipeline that evaluates signal severity, temporal recency, and spatial spillover pressure from neighboring regions.

Users can click any region to access a Civilian Safety Brief summarizing the key drivers behind the risk level and recent alerts.

AI Intelligence Layer

The analytical backbone of StrataWatch is a multi-stage machine learning intelligence engine designed to transform heterogeneous event streams into interpretable risk intelligence.

The system performs:

  • unsupervised anomaly detection to identify statistically irregular clusters of disruption signals
  • graph-based cascade probability estimation to model how infrastructure failures propagate across dependency networks
  • short-horizon time-series forecasting to project regional risk trajectories over the next 24 hours
  • large-language-model intelligence synthesis to convert quantitative outputs into structured situational briefings
  • social signal extraction to detect emerging narratives in live news streams

Rather than treating each dataset independently, StrataWatch performs cross-source feature fusion, allowing correlated signals — such as seismic events, environmental hazards, and geopolitical developments — to reinforce each other when evaluating emerging instability.

Infrastructure Cascade Modeling

StrataWatch models infrastructure as a directed dependency graph composed of interconnected nodes such as power grids, hospitals, transportation hubs, supply chains, and communication networks.

When disruptions occur, the system estimates failure propagation probabilities across the network using graph traversal and probabilistic stress propagation models.

The cascade visualization allows users to observe how disruptions propagate through the network, transforming isolated events into systemic failures.

Simulation Lab

Users can drill down from global analysis to local environments using the Simulation Lab.

Here, clusters of sites and individual buildings can be selected for localized disaster simulations.

The platform generates a simplified 3D building model and runs scenario-specific simulations such as:

  • fire propagation
  • flood water spread
  • earthquake structural stress

Each simulation produces a multi-agent incident analysis, where specialized AI agents independently evaluate the scenario.

The system synthesizes these assessments into a consensus output including:

  • building risk scores
  • evacuation route recommendations
  • structural hazard zones
  • responder entry points

The result is a dynamic simulation environment designed to help users understand how global disruptions translate into local consequences.

How we built it

StrataWatch was engineered as a modular full-stack platform optimized for high-performance geospatial visualization and real-time intelligence synthesis.

Frontend

The interface was built using:

  • Next.js
  • React
  • Tailwind CSS
  • Mapbox GL for geospatial visualization
  • React Three Fiber / Three.js for 3D simulation rendering

The main interface includes a global heatmap, infrastructure network visualizations, and a 3D simulation viewer.

Backend

The backend processes incoming signals and generates risk intelligence.

Technologies include:

  • Node.js API routes
  • Python microservices
  • Redis for state management
  • WebSockets for real-time updates

Data Sources

StrataWatch ingests hybrid signals from both live sources and simulated fallback signals.

Live data includes:

  • USGS earthquake feeds
  • NASA EONET environmental events
  • Open-Meteo weather hazards
  • live conflict and news signals

Simulated signals ensure reliable demos and consistent system activity.

AI and Machine Learning

The analytical engine relies on a layered machine learning architecture designed to transform raw event streams into probabilistic situational intelligence.

Incoming signals are first normalized into a unified event schema before entering a feature engineering pipeline that extracts temporal, spatial, and relational features. These features feed several analytical components including anomaly detection models, infrastructure cascade estimators, and short-term forecasting systems.

To generate interpretable intelligence outputs, the system uses a language model reasoning layer that synthesizes structured analytics into human-readable briefings grounded in the underlying data.

Simulation Engine

The Simulation Lab generates simplified digital twins of selected buildings.

Scenario-specific visual effects simulate environmental dynamics such as:

  • fire and smoke behavior
  • flood water propagation
  • earthquake-induced structural stress

Multiple AI agents independently analyze each scenario and produce structured assessments that are aggregated into a consensus incident report.

Challenges we ran into

One of the most difficult challenges was designing a system that could operate across multiple spatial scales simultaneously.

Global intelligence systems typically focus on large-scale monitoring, while simulation environments focus on localized disaster scenarios. Integrating both into a single responsive interface required careful system design and optimization.

Another challenge was balancing signal sensitivity with interpretability.

Highly sensitive detection systems produce too many alerts, while overly conservative models fail to identify emerging risks. We ultimately optimized the scoring engine to prioritize early detection while contextualizing alerts with explanatory intelligence summaries.

Rendering performance also required extensive optimization. The platform simultaneously displays global maps, infrastructure networks, and interactive 3D simulations, which required careful performance tuning to maintain responsiveness.

Accomplishments that we're proud of

What we are most proud of is successfully connecting global-scale intelligence with local decision-making.

StrataWatch demonstrates that it is possible to combine:

  • global disruption monitoring
  • predictive infrastructure cascade modeling
  • AI-generated intelligence summaries
  • building-level disaster simulations

within a single platform.

Instead of simply visualizing data, the system provides a layered understanding of how risks emerge, propagate, and ultimately affect real-world environments.

What we learned

One of our most important lessons was that data fusion dramatically increases predictive insight.

Individual signals — such as seismic activity, weather anomalies, or news sentiment — rarely provide complete context. However, when these signals are combined into a unified analytical pipeline, patterns emerge that would otherwise remain invisible.

We also learned that the most valuable outputs are not complex models or dashboards, but clear and interpretable intelligence summaries.

Users need actionable context: what is happening, why it matters, and what to watch next.

Designing the system around that principle fundamentally shaped StrataWatch.

What's next for StrataWatch

StrataWatch currently demonstrates the foundation of a global disruption intelligence platform, but its potential extends far beyond this prototype.

Future directions include:

  • integrating satellite imagery analysis
  • expanding predictive disaster forecasting models
  • incorporating additional real-time infrastructure datasets
  • enabling city-scale digital twin simulations
  • building collaboration tools for emergency response teams

Our long-term vision is a system that helps communities anticipate disruptions before they escalate — and respond more effectively when they do.

StrataWatch represents an early step toward a future where situational awareness is shared globally and resilience becomes accessible to everyone.

Share this project:

Updates