Back to Projects
NLP / Civic Tech

Sambodhan

AI grievance redressal system with priority prediction, department classification, and FAQ semantic similarity — FuseMachines AI Fellowship capstone.

Tech Stack

PythonPyTorchScikit-learnNLPSentence TransformersFastAPIReact

Overview

Sambodhan was built as the capstone project for the FuseMachines AI Fellowship 2025. It is an intelligent civic grievance management system that automatically processes citizen complaints: classifying them by the responsible department, predicting urgency priority, and surfacing semantically similar resolved FAQs to help agents respond faster. The system was trained on real civic complaint datasets and deployed as a web application for demonstration.

Key Highlights

  • Multi-label department classification using fine-tuned transformer models

  • Priority prediction model trained on historical grievance resolution data

  • FAQ semantic similarity with sentence embeddings for instant agent assist

  • Full-stack deployment with FastAPI backend and React frontend

  • Built as part of FuseMachines AI Fellowship 2025