McCready Law AI Assistant Case Study

McCready Law AI Assistant

RAG-Powered Legal Case Intelligence Platform

Built an AI-powered case management assistant for McCready Law that pulls criminal case data from their SmartAdvocate system, vectorizes it with Pinecone, and enables attorneys to search and query thousands of cases using natural language through a conversational chat interface.

The Challenge

McCready Law attorneys needed a way to quickly search and query thousands of criminal cases without manually digging through their case management system — they needed answers in seconds, not hours.

Thousands of criminal cases in SmartAdvocate with no intelligent search capability
Attorneys spent hours manually searching for case patterns, precedents, and status updates
No way to ask natural language questions like 'show me all DUI cases from 2024 still in discovery'
Case data was siloed in the CMS with no semantic understanding of case content

Our Solution

SmartAdvocate Data Pipeline

Built automated sync pipelines that pull case data from the SmartAdvocate API, extract 30+ metadata fields per case (status, type, parties, dates, staff, fees), and keep the vector database current with daily syncs.

Vector Embedding & Semantic Search

Vectorized all case data using Pinecone with llama-text-embed-v2 embeddings and bge-reranker-v2-m3 for reranking, enabling semantic search across thousands of cases with metadata filtering by status, type, and year.

LangGraph AI Agent

Built an agentic workflow with LangGraph and Google Vertex AI that autonomously decides which tools to use — semantic search, case lookup, pattern counting, or time-on-desk analysis — based on the attorney's natural language query.

Chat Interface

Developed a React chat UI with real-time streaming, markdown rendering, dark mode, and Firebase authentication — restricted to approved McCready Law email domains.