🌍 Atlas AI

A Professional Multi-Agent System, Digital Twin, and Scenario Simulator

Ready for Submission

📖 Challenge Submission Details

1. Chosen Vertical: Smart Stadium & Event Infrastructure Management

Managing large-scale infrastructural systems like stadiums during live events is highly complex and reactive. Security, crowd flow, environmental controls, and parking are traditionally siloed. Atlas AI targets this vertical by serving as a central, proactive intelligence layer that connects physical infrastructure (via a Digital Twin) with a network of specialized AI agents.

2. Approach and Logic

Our approach bridges Multi-Agent Orchestration with Digital Twin Technology:

• Data Ingestion: The Digital Twin acts as the single source of truth, mocking real-time telemetry (crowd density, temperature, structural health, parking capacity).
• Multi-Agent System: Specialized AI agents (e.g., Crowd Flow Agent, Security Agent, Environmental Agent) continuously monitor their specific domains.
• Centralized Reasoning (Atlas Brain): The agents feed their findings to the "Atlas Brain," which uses the Google Gemini LLM to synthesize the data, identify cross-domain anomalies, and generate high-confidence predictions and recommendations.
• Scenario Simulation: Administrators can inject stress-test scenarios (e.g., "Fire Alarm in Sector B") to observe how the AI predicts crowd panic and recommends evacuation routes in real-time.

3. How the Solution Works

1. The Dashboard: Operators view a real-time Command Center showing active alerts, stadium health, and crowd heatmaps.
2. Dynamic AI Engine: When an anomaly occurs (e.g., sudden crowd density spike), the aiService packages the current Digital Twin state into a strict JSON schema prompt and sends it to the Gemini API.
3. Structured Response: Gemini responds with structured JSON containing:
   • Predictions: What is likely to happen next (e.g., bottleneck at Gate 4).
   • Recommendations: Specific actions (e.g., deploy stewards, open auxiliary exits).
   • Reasoning: Step-by-step logic explaining why the AI made these recommendations, ensuring full transparency.
4. Execution: The recommendations are displayed beautifully in the UI for the human operator to approve.

4. Assumptions Made

• We assume that the physical stadium is heavily instrumented with IoT sensors (cameras, temperature sensors, ticketing gates) capable of feeding real-time data to the Digital Twin.
• For the sake of this prototype, sensor telemetry is simulated dynamically on the frontend.
• We assume human-in-the-loop operation: the AI recommends, but a human operator ultimately approves critical actions.

---

🛠 Tech Stack & Architecture

• Frontend Framework: React 19 + Vite (TypeScript)
• Styling & Animations: Vanilla CSS + Framer Motion (for premium micro-animations)
• AI Integration: @google/genai SDK for robust, dynamic reasoning
• State Management: React Context API for global simulation state sync

🚀 Installation & Setup

1. Clone the repository:

   git clone https://github.com/Hsisdie/Atlas-ai.git
   cd Atlas-ai

2. Install dependencies:

   npm install

3. Configure Environment Variables: Create a .env file in the root directory and add your Gemini API key:

   VITE_GEMINI_API_KEY=your_gemini_api_key_here

4. Start the Application:

   npm run dev

   The application will be available at http://localhost:5173.

🔒 Security Note

This repository follows security best practices. The .env file is excluded via .gitignore to ensure API keys are never exposed in version control. If deploying to production, environment variables should be injected securely via the hosting provider.

🤝 Team

• Developed for the 2026 AI Challenge
• License: MIT