Legis Synapse

Inspiration

The idea for **Legis Synapse** sparked from a simple observation: legal documents are scary for most people. We saw our friends and family get confused and stressed out by things like rental agreements or their first job contracts. It felt unfair that they either had to pay a lot for a lawyer or just sign and hope for the best. We learned about this huge "access to justice" gap, where everyday people are left to navigate complex legal jargon on their own. We were inspired to use the power of **GenAI** to build something that could level the playing field and give everyone the confidence to understand the documents that shape their lives.

What it does

So, we built Legis Synapse, our AI-powered "Legal Co-Pilot" designed to help you out! It's not just another PDF viewer; it's a smart assistant that: Translates Legalese: It takes all that complicated legal language and turns it into simple, plain English that anyone can understand. Spots Risks for You: It automatically flags clauses that might be risky, non-standard, or don't comply with local laws, so you know exactly what to look out for. Simulates Negotiations: This is the coolest part! It uses a "Digital Twin" simulation to predict how a negotiation might go, helping you avoid arguments and disputes before they even start. Makes it Official (and Secure!): It can generate a blockchain smart contract from the final document, making it super secure. Speaks Your Language: We included multi-language voice support, so it's accessible and easy to use for people all across India.

How we built it

We dove into the **Google Cloud** ecosystem to bring **Legis Synapse** to life. The "brain" of our project is a swarm of **24 AI agents** that we orchestrated using **Vertex AI Agent Builder**. We chose **Gemini Pro** for the core reasoning because of its powerful capabilities. To make our AI smart about local laws, we fed it information from a **Vector Database** and a **Legal Knowledge Graph**. For the backend, we went serverless with **Google Cloud Run** to keep it fast and scalable, and we used the **ADK (Agent Dev Kit)** to build our user dashboard. It was a sprint, but we put together a full-stack, multi-layer architecture we're really proud of.

Challenges we ran into

We quickly realized this wasn't just a coding project. Our biggest challenge, by far, was something we hadn't thought much about at first: the **Unauthorized Practice of Law (UPL)**. It was a huge challenge to design the app to be genuinely helpful without crossing the line into giving "legal advice," which is illegal. On the technical side, we spent a lot of time wrestling with AI "hallucinations". An AI making up facts is a funny curiosity in some apps, but in a legal tool, it's a disaster waiting to happen, so we had to build robust systems to prevent that.

Accomplishments that we're proud of

We're incredibly proud of building a tool that feels like it has a real superpower. The Digital Twin Simulation felt like something out of science fiction at first, but we actually designed a system for it that can predict negotiation outcomes. We are also proud of building a **"Legal Co-Pilot"** powered by a complex **24-agent swarm**, which is a big step up from simple analysis tools. The fact that our system is designed to reduce legal review time by 90% and disputes by 60% is an accomplishment we're excited about, as it shows the real-world impact our project can have.

What we learned

This project taught us so much more than just how to use new APIs. The biggest lesson was that technology, especially AI, has to be built responsibly. We learned first-hand the importance of technical safeguards like a **Retrieval-Augmented Generation (RAG)** pipeline to keep the AI grounded in facts and prevent it from making things up. We also learned that in a field like law, user trust is everything. That's why we made it a priority to build in transparency, like making sure every summary the AI generates can be clicked to highlight the original text in the document.

What's next for Legis Synapse

We feel like we're just getting started! Our dream is for **Legis Synapse** to become the go-to tool for anyone facing a confusing document. Next, we want to expand the types of documents it can handle, like freelance contracts and more job offers. We're also planning to build out even more powerful features, like a tool that lets you compare two versions of a document to see what's changed. Ultimately, we want to add an AI negotiation assistant that can suggest better wording for clauses, truly making it a comprehensive companion for our users.

Built With

• adk
• adk-(agent-dev-kit)
• adk-dashboards
• ai-&-agent-layer:-gemini-pro
• and-google-translate-api.-data-&-knowledge-systems:-google-cloud-storage
• and-legal-knowledge-graphs-(neo4j).-infrastructure-&-backend:-google-cloud-run
• blockchain
• docker
• dotenv
• firestore
• gemini-api
• gemini-pro
• google-cloud
• google-cloud-run
• google-cloud-vision-api
• google-translate-api
• python
• vector-databases-(faiss/pinecone)
• vertex-ai-agent-builder