Project 04
Eye of Horus – SPARKS
2026AI SystemsFull-Stack · Crowd Intelligence
Eye of Horus – SPARKS is a full-stack incident prediction platform designed to monitor live events and estimate crowd risk in real time. It ingests heterogeneous signals across social media, environmental feeds, and behavioral data, aggregates them into a unified risk model, and uses large-scale agent simulation to anticipate dangerous patterns before they surface. The system combines careful engineering with LLM-driven reasoning to turn noisy data into operational insight for organizers and safety teams.
A real-time crowd intelligence system that fuses live signals and agent simulations to predict incident risk at large-scale events.
Role
System architecture · Backend & frontend engineering · AI integration
Stack
- Python
- FastAPI
- React.js
- PostgreSQL
- Claude API
Highlights
- —Engineered a real-time sentiment and signal aggregation layer processing 40+ channels across social, environmental, and behavioral data streams
- —Built a 10,000-agent distributed swarm simulation (“Oracle” layer) powered by Claude API to model emergent crowd dynamics at event scale
- —Achieved 92.7% accuracy in backtesting on real events including Astroworld, Coachella, and the Super Bowl
- —Shipped a 49-file codebase with 21/21 tests passing, structured for production-ready deployment
- —Positioned as a safety-focused platform aimed at helping organizers and cities anticipate, visualize, and mitigate crowd incidents before they escalate