ai-document-analyst

📄 AI Document Analyst

Reads contracts, employment agreements, NDAs, and investor documents. Extracts every key clause, flags risky language, and gives you a plain-English summary — in 30 seconds.

What It Does

Upload any legal document. The agent reads it like a senior attorney, tells you what’s in it, what’s dangerous, and what you should negotiate before signing.

Pipeline: PDF/Text Input → Type Detection → Clause Extraction → Risk Scan → Executive Summary → Q&A Interface

Real-World Impact

• $300–$500/hr attorney time replaced for first-pass review
• Catches hidden traps: perpetual non-competes, IP grabs, one-sided termination
• Makes legal documents understandable for non-lawyers

Tech Stack

| Tool | Role | |——|——| | OpenAI GPT-4o | Document reading + clause extraction | | GPT-4o Vision | Image/scanned PDF analysis | | PyPDF2 | PDF text extraction | | Python | Pipeline orchestration |

Setup

pip install openai PyPDF2
export OPENAI_API_KEY=your_key

# Analyze a file
python agent.py path/to/contract.pdf

# Or use the demo (built-in sample contract)
python agent.py

Document Types Supported

• Employment Contracts
• SaaS Agreements
• NDAs (Non-Disclosure Agreements)
• Investor/Term Sheet Agreements
• General Contracts

Sample Output

🔴 [CRITICAL] IP Assignment Beyond Employment
   Why it matters: All inventions — even personal projects on weekends —
   become company property. This is unusually broad and should be removed.
   Action: Negotiate to limit IP assignment to work directly related to
   company business during working hours only.

Interactive Q&A

After analysis, ask any question about the document in plain English:

Your question: Can I work for a competitor after leaving?
Answer: The non-compete clause (Section 4) prohibits you from working in
any competitive business for 5 years after termination, anywhere in the US.
This is extremely broad and likely unenforceable in many states.

---

Built by Philip | AI Agents & Automation Specialist