I, Syeda Ameena, Suyash Gupte and Suman P participated in the Google Agentic AI Day 2025, Bengaluru.
Out of 57,000+ developer and 9,100+ idea submissions, our team got shortlisted for Google Cloud Agentic Al Day 2025, powered by Hack2skill. We built a multi-agent system with Google ADK on the use case Improving safety at large public events in a 30-hour in-person Al sprint.We presents Drishti, an AI-driven situational awareness and enforcement platform aimed at enhancing public safety at large-scale events in India. The system transforms passive video surveillance into active safety management by combining real-time video analytics, predictive AI, and digital accountability.
Drishti tackles the unique safety challenges of high-density public gatherings like rallies, festivals, and marathons in India. Instead of relying solely on human supervision, it implements a closed-loop model of real-time detection, automated intervention, and behavioral enforcement. By integrating technologies such as Google Cloud’s Vertex AI, Gemini, and digital ID systems, the platform becomes a self-regulating mechanism for crowd safety and order.
The solution is structured around five core components:
Predictive Analytics: Real-time crowd congestion prediction using video feeds and AI forecasting.
AI-Powered Dispatch: Automated emergency response leveraging GPS and live congestion data.
Digital Accountability: Tracking attendee behavior via digital IDs (QR codes, phone numbers, Aadhaar), and issuing nudges or post-event penalties.
Scalable Multi-Agent System: Modular agents (Predictive Agent, Dispatch Agent, Lost & Found Agent, etc.) working independently in parallel.
Cultural Adaptation: Designed for India’s event complexity—low compliance, high density, digital familiarity.
Key Features:
• Natural language querying for field commanders.
• Context-aware summaries and predictions with Gemini.
• Lost & Found via facial recognition and drone assistance.
• Real-time rule enforcement via mobile-linked identities.