• Home
  • About
    • LiaX photo

      LiaX

      Running in my time zone

    • Learn More
    • LinkedIn
    • Github
  • Posts
    • All Posts
    • All Tags
  • Projects

All Projects

A Retrieval-Augmented Legal Intelligence System

This is a Retrieval-Augmented Generation (RAG) system designed to provide accurate and context-aware legal assistance for German law. The platform integrates statutory legislation from the German Bundestag and real-world court decisions from OpenLegalData through a hybrid retrieval architecture.

By combining multilingual embeddings, vector search, LLM-based query classification, and conversational memory, the system delivers grounded legal answers while significantly reducing hallucinations commonly observed in standalone language models.

Python LangChain LangGraph ChromaDB GPT-4o-mini Django Docker
  • Hybrid retrieval across German statutory laws and legal cases
  • LLM-based query classification with 90.6% precision
  • Multilingual legal embeddings using multilingual-e5-large-instruct
  • Conversation memory powered by LangGraph
  • Retrieval-grounded generation to reduce hallucinations
  • Django-based interactive legal assistant interface
Your browser does not support the video tag.