AI-Powered Legal Citation Analyzer
Client capstone with Wolters Kluwer: Agentic RAG system using Neo4j knowledge graphs and multi-LLM orchestration to automate legal precedent research.
The Problem
Legal Research Bottleneck
Attorneys spend hours manually verifying if legal precedents remain valid ("Good Law"). Wolters Kluwer needed an AI-powered solution to automate citation analysis and reduce research time.
The Approach
Knowledge Graph + Agentic RAG
Built an agentic RAG system with Neo4j knowledge graphs storing 3,500+ ADA cases and 5,500+ citations. Orchestrated multi-LLM pipelines (Claude 3.5 Sonnet, Llama 3-70B) using LangChain and LangGraph on AWS Bedrock, with a Streamlit interface.
Technologies & Methods
The Results
Production MVP Delivered
Delivered a dual-mode application with Case Lookup and Chatbot interfaces, plus a configurable case-labeling system that lets users align outputs with their judgment and risk tolerance. Achieved 67% citation classification accuracy using a multi-LLM ensemble with paragraph-level rationales.