Prae8 Legal Search

The problem

Legal research costs small practices time and money. Traditional legal databases are expensive and hard to learn. Small firms and solo lawyers need an affordable, simple way to search case law, statutes and precedents.

What I built

I co-founded Prae8 and built an MVP: a search tool that answers questions over a lawyer's own documents and cites its sources.

• A RAG pipeline that retrieves the right passages and generates answers with sources
• ChromaDB for vector search and Hugging Face embeddings for legal text
• FastAPI backend and Next.js frontend
• Open-source models on private Runpod servers: documents stay in our environment and train no outside model

Every lawyer I met asked the same question: where does my data go? Privacy shaped the design, and I moved from commercial APIs to self-hosted models.

What I learned

A lawyer tested the MVP and needed good prompts more than a complex pipeline. Complex is not always better.

Stack

Python • ChromaDB • Hugging Face • RAG pipeline • FastAPI • Next.js • Runpod