🛡️ Contract Simplifier — AI-Powered Contract Risk Analysis

🌍 Inspiration

Business contracts often carry hidden risks that can affect compliance, finances, or security — yet reviewing them manually is slow, subjective, and prone to oversight. Our team wanted to build an AI system that helps organizations evaluate vendor risk before signing a deal, in real time. We were inspired by the growing use of AI in due diligence and compliance automation and wanted to make risk intelligence accessible even to small teams without requiring a dedicated legal department.

🧠 What We Built

Contract Simplifier is a real-time, AI-powered risk analysis system that identifies, evaluates, and visualizes potential vendor risks before an organization enters into a business contract.

It combines LangGraph, ChromaDB, Bright Data, FastAPI, and React to form an intelligent end-to-end pipeline:

User enters a query such as “Analyze Tesla with high criticality.”
The backend orchestrates multiple components:

LangGraph controls the AI workflow as a state machine.
Bright Data scrapes credible company data in real time.
ChromaDB stores and retrieves context embeddings.
OpenAI GPT-4 analyzes risk factors and scores vendors.
1. The React frontend displays a live-streaming visualization of each analysis stage, showing how data moves from collection to insight.

This flow allows users to see the reasoning behind AI-driven conclusions, increasing transparency and trust.

⚙️ How We Built It

We followed a modular micro-frontend and microservice approach:

Backend (FastAPI + LangGraph):
- Designed a pipeline with multiple “nodes” for classification, scraping, validation, and LLM evaluation.
- Implemented streaming responses via Server-Sent Events (SSE) for real-time updates.
- Integrated ChromaDB as a semantic memory layer for contextual recall.
Frontend (React + Tailwind CSS):
- Developed a split-screen interface — chat panel on the left, risk visualization on the right.
- Used animated progress indicators to show each step of the risk workflow.
- Rendered color-coded risk scores with smooth transitions and responsive layout.
Data Layer:
- Leveraged Bright Data for trusted source collection.
- Applied domain-based filtering to prioritize credible sites (e.g., .gov, .edu, Bloomberg, Reuters).

🔍 Key Features

AI Risk Scoring across four critical dimensions:
- 🏦 Financial Risk
- 🔐 Security Risk
- 📉 Reputation Risk
- 🧩 Resilience Strength
Streaming Visualization: View live updates of the analysis pipeline.
Source Verification: Automatically filters unreliable or duplicate data.
Vector Search with ChromaDB: Enhances context relevance for LLM queries.
Smart Classification: Detects whether the user request is a risk query or general chat.

🧩 Architecture Overview

┌─────────────┐         ┌──────────────┐         ┌─────────────┐
│   React     │ ◄─SSE──►│   FastAPI    │ ◄──────►│  LangGraph  │
│   Frontend  │         │   Backend    │         │   Pipeline  │
└─────────────┘         └──────────────┘         └─────────────┘
                               │                         │
                               ▼                         ▼
                        ┌──────────────┐         ┌─────────────┐
                        │  Web Scraper │         │  ChromaDB   │
                        └──────────────┘         └─────────────┘

💡 What We Learned

AI pipelines benefit from clear state management. LangGraph helped us structure complex workflows that involved multiple data sources and LLM reasoning stages.
Streaming enhances user trust. Showing every step of analysis makes AI decisions more explainable and engaging.
Data credibility is crucial. Many scraped sources are unreliable, so integrating domain verification was key.
Frontend performance tuning matters. Handling SSE in React required careful optimization to avoid UI lag during real-time updates.
Cross-team collaboration is smoother with modular design. Each team member could iterate independently on backend, frontend, or AI pipeline.

🧠 Challenges We Faced

Latency management: AI and web scraping operations can be slow, so we used asynchronous FastAPI tasks and caching.
LLM consistency: Different runs could produce slightly varied risk ratings, so we implemented post-processing checks.
Data validation: Ensuring only credible, non-duplicate sources reached the model required multiple iterations of our filtering logic.
Frontend synchronization: Coordinating the live streaming events with changing backend states was initially tricky but solved using an event-driven architecture.

🚀 Impact

The system demonstrates how AI can act as an assistant for contract risk evaluation — performing tasks that would normally take analysts hours. By providing real-time transparency, contextual memory, and automated reasoning, Contract Simplifier has the potential to help organizations reduce decision-making time and improve compliance readiness.

🧰 Tech Stack

Layer	Technology
AI Workflow	LangGraph, LangChain
Backend	FastAPI (Python 3.8+)
Frontend	React 18+, Tailwind CSS
Vector Database	ChromaDB
Web Scraping	Bright Data API
LLM Engine	OpenAI GPT-4
Streaming	Server-Sent Events (SSE)

🧪 Future Enhancements

Batch contract analysis for multiple vendors
Integration with CRM or ERP systems
Historical risk trend tracking
PDF report export
Multilingual support
Custom risk factor weighting

🤝 Team

Sanjay Sakthivel — AI & Backend (LangGraph, ChromaDB, FastAPI)
Aman Nindra — Frontend & Streaming & Database Integration (React, Tailwind, Firebase, AWS)
Rajbir Longia — Web Scraping & API Integration (Bright Data, REST)

❤️ Closing Thoughts

We built Contract Simplifier to show how AI can transform due diligence from a static checklist into a dynamic, explainable, and data-driven process. It’s not just about predicting risk — it’s about understanding it in real time.