The Indian legal system has always been based on a large amount of complex legal documents, laws, and case rulings, making it hard for individuals to understand and navigate.

Inspired by the latest developments in Large Language Models (LLMs) space. I've finetuned several well-known Large Language Models on Indian laws and regulations, relevant articles, legal advice forum threads, and other legal text corpus using QLoRA technique. This finetuned models include Meta's LLaMA-3-8b and LLaMA-2-7b, Mistral AI's Mistral-7B model and other popular LLM models. These models, including their quantized versions, are available on my Hugging Face profile. LawSage also includes Retrieval-Augmented Generation (RAG) based chat interface. This chat interface uses a dataset that includes the Indian Constitution, Central Acts, and State Acts legislation, which I compiled into a text corpus and stored as vector embeddings for retrieval to use them as underlying context for generating responses.