Back to Projects
Machine Learning
Legal Information Retrieval System.
Executive Summary
Developed a legal question-answering system using a custom-trained LegalBERT model integrated with Retrieval-Augmented Generation (RAG) techniques.
Background & Implementation
This system focuses on high-precision information retrieval across over 5000 Indian case judgments and statutes including the RTI Act. By fine-tuning LegalBERT for semantic matching and utilizing LlamaIndex for RAG, the system achieves superior performance in identifying relevant legal precedents compared to standard LLM retrieval.
Key Technical Specs
Architecture
LegalBERT + RAG
Date
Feb - May 2025
Stack
PythonTransformersLlamaIndexFastAPI