Lawly AI

Inspiration

Legal problems hit people when they’re stressed, confused, and least able to navigate a broken system. Most Americans choose lawyers based on ads, Google rankings, or whoever replies first not on actual case history. We wanted to fix that and give everyone fair representation, powered by real data instead of marketing.

What it does

Our platform reads and interprets real court opinions, matches users with attorneys who have proven results in similar cases, and handles all the back-and-forth emails to negotiate a fair agreement. Users get clarity, the right lawyer, and a streamlined path to legal help.

How we built it

We built a full pipeline with a Next.js frontend and a FastAPI backend running on AWS (Lambda, S3) that orchestrates OpenAI, Gemini, and LangGraph for legal opinion understanding and workflow control. Our custom matching engine uses BM25, Apollo, PostgreSQL, and Supabase to store and rank attorneys by real case performance, with Pandas and NumPy handling analysis and scoring. On top of that, we integrated automated email threading to remove the tedious back-and-forth of scheduling.

Challenges we ran into

Parsing dense legal opinions at scale, reliably identifying case topics, and matching them to attorney performance proved extremely complex. Automating live email threads without breaking formatting or causing confusion required multiple iterations. Balancing accuracy, latency, and user simplicity was another major hurdle.

Accomplishments that we're proud of

We built a fully functional, end-to-end system in a short timeframe from case analysis to attorney matching to automated email conversations. The accuracy of our case-to-lawyer matching far exceeded our expectations, and our live email threading feature created a seamless user experience.

What we learned

We learned how messy and fragmented legal data is, how difficult it is for everyday people to understand their options, and how powerful LLMs can be when combined with structured pipelines. We also learned how important simplicity is when building tools for high-stress situations.

What's next for Lawly AI

We plan to expand to more jurisdictions, train the system on additional court datasets, integrate document upload/analysis, and build a full consultation dashboard. Long term, we aim to make legal expertise accessible, transparent, and data-driven for everyone.