Clause

Architecture Diagram
Interactive Multilingual Chatbot
Primary Analysis and Reccomendation
Pre-Document Upload Information
Document Upload
Contract Highlighting and Annotation
Chrome Extension Demo

Inspiration

Information asymmetry is weaponized against ordinary people. Billions of dollars in tenant rights, medical bill errors, and consumer protections go unclaimed because they're buried in fine print.

When surveyed, renters overwhelmingly answered "Don't Know" when asked about their legal rights regarding security deposits, including basic protections like how many days landlords have to return deposits. Massachusetts law (M.G.L. c. 186, § 15B) entitles tenants to triple damages for security deposit violations (Mass.gov), yet most never collect because they don't know their rights or find the legal process intimidating.

The problem extends beyond housing: an estimated 80% of medical bills in the United States contain errors (BillFlash) (Dialog Health), and 45% of insured adults received a medical bill for a service they believed should be covered by their insurance (AJMC). Yet 54% of respondents who didn't contest billing issues were unaware they had the right to do so, with this knowledge gap being most pronounced among younger adults. (AJMC)

We built Clause to level the playing field, turning legal complexity into instant, actionable enforcement.

What it does

Clause is an AI-powered legal enforcement platform. Upload a lease, medical bill, or contract! Our system detects violations, calculates damages, and generates demand letters with statute citations. A Chrome extension flags illegal clauses in real-time on Airbnb listings. Everything is timestamped on Solana's blockchain for tamper-proof evidence in court. Ask questions via voice or text (in any language), and get instant answers grounded in Massachusetts law.

How we built it

Frontend: React web app + Chrome extension for real-time scanning
Backend: FastAPI with PII redaction (spaCy NER + regex) before processing
RAG Pipeline: Snowflake Cortex (vector embeddings via snowflake-arctic-embed-l-v2.0) + chunked legal documents stored in Snowflake tables
LLM: Gemini 2.0 Flash Thinking for legal analysis, grounded by retrieved statutes
Blockchain: Solana for immutable document hashing + future escrow/reputation features
Storage: Cloudflare R2 for documents, Solana for metadata
Deployment: Vultr Cloud hosting

See architecture diagram for full data flow.

Challenges we ran into

1) LLM hallucinations: Gemini would cite non-existent statutes. Fixed by strict RAG grounding, only generate answers from retrieved chunks, never from model knowledge alone.

2) PII security: Leases contain SSNs and addresses. Built redaction pipeline to strip PII before analysis while maintaining encrypted mappings for demand letter generation.

Accomplishments that we're proud of

1) Built end-to-end RAG pipeline from scratch using Snowflake Cortex.

2) Achieved zero hallucinations in legal analysis through strict retrieval-grounded generation

3) Created real-time contract scanning via Chrome extension that works on live Airbnb pages

4) Implemented blockchain timestamping for court-admissible evidence

5) Designed PII-safe architecture that never exposes sensitive data to external APIs

6) ElevenLabs voice chat and translation for accessibility to vulnerable populations

7) Compiled a growing policies library consisting of MA rental, medical, and consumer law corpus

What we learned

1) Snowflake's vector search capabilities and how to optimize similarity queries with 1024-dim embeddings

2) The critical importance of citation grounding when LLMs handle high-stakes domains like law

3) How to architect RAG systems that balance context window limits with comprehensive legal coverage

4) Blockchain isn't just hype! Immutable timestamps create real legal value for evidence chains

5) User trust requires transparency. Showing source statutes alongside AI answers is non-negotiable

What's next for Clause

1) Neuro-Symbolic AI: Integrate Stanford Law's research on combining neural models with symbolic reasoning engines. This hybrid approach would encode legal logic rules (e.g., "IF deposit > 1 month's rent AND withheld > 30 days THEN triple damages apply") alongside LLM analysis, eliminating hallucinations while maintaining natural language understanding.

2) Smart Contract Enforcement: Fully automate escrow → demand letter → small claims filing → settlement distribution via Solana programs.

3) Landlord Reputation NFTs: Mint violation records as public blockchain tokens tied to property addresses, creating economic incentives for compliance.