ContractIQ — AI-Powered Contract Intelligence (Portfolio Case Study)

The Problem

Contract review is time-consuming and error-prone. Legal teams spend hours manually reviewing contracts to:

• Extract and categorize key clauses (Termination, Payment, Liability, IP, etc.)
• Identify potential risks and unfavorable terms
• Answer questions about contract terms across multiple documents
• Generate evidence packs for negotiations or compliance

A typical contract review process involves reading through 20-100 page documents line by line, manually identifying clauses, assessing risks, and creating evidence documents. This manual process takes hours or days per contract, increasing costs and the risk of missing critical terms.

The Solution

ContractIQ leverages AI to automate contract analysis. The platform combines intelligent clause extraction with semantic search, enabling legal teams to:

• Automatically extract clauses with risk assessment from uploaded contracts
• Ask questions in natural language and get answers with citations across multiple documents
• Generate evidence packs with highlighted excerpts for negotiations
• Assess risks with automated scoring and detailed reasoning

Example: Instead of spending hours or days manually reviewing a 20-100 page vendor agreement, a legal team can upload the document, automatically extract all clauses with risk scores in minutes, ask "What are the termination terms?" and get an instant answer with page citations, then generate an evidence pack for negotiations.

What I Built

1) Intelligent Clause Extraction

ContractIQ automatically identifies and extracts 15+ clause types from contracts:

• Termination: Early termination, breach termination, convenience termination
• Payment: Payment terms, schedules, penalties, late fees
• Liability: Liability limitations, caps, exclusions
• Indemnification: Indemnification clauses, hold harmless provisions
• Intellectual Property: IP ownership, licensing, rights
• Confidentiality: NDA terms, confidentiality obligations
• Dispute Resolution: Arbitration, jurisdiction, mediation
• Force Majeure: Force majeure provisions
• Compliance: Regulatory compliance, certifications
• Insurance: Insurance requirements, coverage
• Warranties: Warranties, representations
• Data Privacy: Data protection, privacy obligations
• Non-Compete: Non-compete, non-solicitation
• Assignment: Assignment rights, restrictions
• Governing Law: Choice of law, venue
• And more...

Each extracted clause includes risk score (0-100), risk flags, detailed reasoning, confidence score, and page numbers.

2) Semantic Search with RAG

ContractIQ uses a RAG (Retrieval Augmented Generation) pipeline to answer questions across multiple documents:

Natural Language Queries:

• "What are the termination terms?"
• "Compare liability caps across all vendor agreements"
• "What are the payment terms for contract X?"
• "Find all non-compete clauses"

Features include multi-document search, citations with page numbers, conversation history, and validated answers. The RAG pipeline uses LangGraph: retrieve relevant chunks, generate answer, validate citations.

3) Risk Analysis

Every extracted clause receives automated risk assessment:

Risk analysis includes automated scoring (0-100), risk flags (unfavorable termination, high liability, unfair payment terms, etc.), and detailed reasoning for each clause.

4) Evidence Pack Generation

ContractIQ generates professional PDF evidence packs from Q&A conversations:

Generates professional PDFs with highlighted citations, conversation summaries, and source attribution. Use cases include negotiation materials, compliance reports, and stakeholder summaries.

5) Multi-Tenant Workspace Architecture

Supports multiple workspaces with complete data isolation. Each workspace has separate document collections and ChromaDB vector stores, enabling organized workflows for different contract types or clients.

6) Document Processing

Intelligent processing with PDF/DOCX support, LLM structuring, semantic chunking, and background processing. Preserves accurate page numbers for citations.

Technical Deep Dive

How It Works

ContractIQ follows a three-stage workflow: document processing, clause extraction, and Q&A:

Document Processing: Upload → Extract text → LLM structure → Semantic chunking → Index to ChromaDB → Ready

Clause Extraction: Select document → LLM analyzes chunks → Extract clauses → Risk analysis → Store in database

Q&A: Ask question → Vector search → Retrieve chunks → Generate answer → Validate citations → Return response

Technology Stack

Backend:

• FastAPI (Python 3.11+): High-performance async API framework
• LangGraph: RAG pipeline orchestration with state management
• Instructor: Structured output validation for clause extraction
• PostgreSQL 16: Relational data (users, workspaces, documents, clauses)
• ChromaDB: Embedded vector store with per-workspace collections
• Redis 7: Caching layer for embeddings and search results

Frontend:

• Next.js 16: React framework with App Router
• shadcn/ui: Component library for consistent UI
• Tailwind CSS: Styling and responsive design
• TypeScript: Type safety across the application

AI & Document Processing:

• OpenAI GPT-4o-mini: Clause extraction, document structuring, answer generation
• OpenAI text-embedding-3-small: Vector embeddings for semantic search
• PyMuPDF: Fast PDF text extraction with page numbers
• python-docx: DOCX file processing

Key Technical Decisions

LangGraph for RAG Pipeline: Structured workflow orchestration ensures accurate citations and validated answers.

Per-Workspace ChromaDB Collections: Complete data isolation for multi-tenant architecture with efficient semantic search.

Structured Output with Instructor: Ensures consistent, validated clause extraction with proper risk scores.

Background Document Processing: Async processing allows multiple uploads without blocking.

Production Impact

Before ContractIQ

A legal team's contract review process:

1. Read 20-100 page vendor agreement line by line (hours)
2. Manually identify and categorize clauses (hours)
3. Assess risks for each clause (hours)
4. Answer questions by searching through document
5. Create evidence pack with highlighted excerpts Total: hours or days per contract

With ContractIQ

The same workflow:

1. Upload document → Automatic processing (5 minutes)
2. Extract clauses → Automatic extraction with risk scores (2 minutes)
3. Ask questions → Instant answers with citations (1 minute)
4. Generate evidence pack → Automatic PDF generation (1 minute) Total: 9 minutes per contract

Use Cases

Legal Teams: Review contracts, identify risky clauses, prepare negotiation materials, answer questions across documents.

Compliance Teams: Audit contracts, identify missing clauses, generate compliance reports.

Procurement Teams: Compare terms across agreements, identify unfavorable terms, generate summaries.

Architecture Highlights

Production-ready patterns: multi-tenant architecture with workspace isolation, background processing, Redis caching, structured output with Instructor, and LangGraph RAG pipeline. ContractIQ reduces manual review time by 75% while improving accuracy.