Inspiration

In the world of tax litigation, one of the biggest challenges is deciding whether a case is worth appealing. Every unnecessary appeal drains public resources, clogs the judicial system, and delays justice.

What it does

•⁠ ⁠Outcome Prediction: Uses machine learning trained on past judicial decisions to predict the likelihood of success in higher courts. •⁠ ⁠Legal Reasoning Engine: Analyzes relevant case law, statutory provisions, and tribunal rulings to generate case-specific legal arguments. •⁠ ⁠Forecasted Opinions: Generates a sample higher court judgment, outlining likely reasoning for upholding or dismissing the appeal. •⁠ ⁠Appeal Optimization: Offers data-backed recommendations to reduce frivolous appeals, enabling smarter closure decisions.

How we built it

We’re building an AI-powered Tax Litigation Co-Pilot — a decision intelligence tool that empowers tax officers with real-time, objective, and legally-grounded recommendations on whether to pursue appeals. We used the following tech stack

  • Python
  • Flask
  • Open AI
  • Milvis DB
  • React
  • Vercel

Challenges we ran into

  • How to upload large PDFs and generate embeddings
  • First time working with a lot of new tech like Vector database and embeddings. So involved a bit of learning curve

Accomplishments that we're proud of

  • Learned a lot of new tech

What we learned

  • RAG
  • Python

What's next for Taxation Appeals Copilot for Lawyers

  • Turn this into a chatbot and make it more interactive
  • Add a voice agent to interact
  • Introduce cloud blob storage for files
  • Implememt this with GraphRAG

Built With

Share this project:

Updates