Скачать или смотреть Smart India Hackathon 2025: Multimodal RAG System for Offline Knowledge Retrieval

Smart India Hackathon 2025 Project – Multimodal RAG System (Team AGLI)

This video presents our project submission for the Smart India Hackathon 2025: a Multimodal Retrieval-Augmented Generation (RAG) System designed to function efficiently in offline environments.

The core idea of our solution is to create a unified semantic retrieval framework capable of ingesting, indexing, and querying data across multiple formats such as documents (DOCX, PDF), images, and audio recordings. Instead of being limited to a single mode of search, the system supports cross-modal queries, enabling text-to-image, image-to-text, audio-to-text, and document-based retrieval within one interface.

The system includes a simple query interface where users can type or speak their queries in natural language. Every result provided is backed by transparent citations, allowing users to trace information back to its original source for accuracy and trust. This makes the solution not only user-friendly but also reliable for critical use cases.

From a technical perspective, the solution leverages Large Language Models (LLMs), optimized vector databases, and multimodal embeddings. Performance challenges such as handling large datasets offline are addressed through chunking, incremental indexing, and deploying lightweight quantized LLMs with GPU acceleration.

The project has wide-ranging benefits:

Government officials can access citation-backed, cross-modal information for informed decision-making.

Healthcare professionals can securely retrieve and analyze medical records, images, and dictated notes.

Students and researchers can consolidate study material and lectures for faster, more accurate referencing.

General users benefit from improved efficiency, transparency, and secure offline information access.


This solution is particularly valuable in regions with limited internet connectivity, such as parts of Northeast India, where reliable access to verified information can directly improve decision-making, healthcare outcomes, and learning experiences.