Urja Setu

AI-Driven Power Utility Feedback System

Problem Context

Citizens reporting electrical hazards — broken poles, sparking transformers — rely on outdated, manual processes. Traditional systems lack urgency detection and geographic precision, leading to delayed response and unresolved safety risks.

The Solution: Headless Architecture

Urja Setu modernizes how citizens report electrical hazards in Gujarat. Instead of a monolithic app, we built a decoupled, headless architecture with three distinct layers:

Citizen App: A lightweight, mobile-first web interface for quick hazard reporting with image uploads.
Technician Dashboard: A separate portal for ground staff to view assignments, map locations, and manage tickets.
Intelligent Backend: A Django REST API that processes uploaded images using YOLOv8 to identify the asset (e.g., "Transformer") and auto-prioritize tickets by severity.

AI Integration

The system uses YOLOv8 (Object Detection & Classification) to analyze citizen-uploaded images. When a report is submitted, the backend automatically triggers an analysis task that identifies the type of electrical asset, detects visible damage, and assigns a severity priority to the ticket — all without manual intervention.

Tech Stack

Django REST Framework YOLOv8 JavaScript (Vanilla) SQLite / PostgreSQL OpenStreetMap

Validation & Results

SSIP Vikas Saptah Hackathon 2025 (6th Edition) — Selected as a State Level Finalist, ranking in the Top 50 teams out of 4,000+ entries across Gujarat. The platform was built to address a real-world governance challenge in public power utility feedback.

Outcome & Learnings

Urja Setu demonstrated that decoupled architectures paired with Computer Vision can transform traditional civic processes. The project highlighted the importance of clean API design, real-time geocoding with OpenStreetMap, and the practical deployment challenges of AI models in resource-constrained environments.