Smart Document Analyzer 📄🤖

An AI-powered document-processing system built using Amazon Textract, AWS Lambda, Amazon Bedrock/SageMaker, and a serverless cloud-native architecture.
This application extracts text from PDFs/images, generates intelligent summaries, and enables question-answering using Retrieval-Augmented Generation (RAG).

---

🚀 Features

🔍 Intelligent OCR

• Upload PDFs or images
• Extracts structured text using Amazon Textract
• Supports multi-page documents

🧠 AI Summaries

• Automated text summarization
• TL;DR bullet generation
• Powered by Amazon Bedrock or SageMaker JumpStart models

❓ Ask-Anything Q&A

• Ask natural-language questions about the uploaded document
• Uses RAG (chunking + embeddings + context retrieval)
• Generates accurate responses using LLMs

☁️ Serverless Architecture

• No servers to manage
• Highly scalable
• Low-cost for student projects

---

🏗️ Architecture

                 ┌──────────────────────────┐
                 │        Frontend          │
                 │  (React + Amplify)       │
                 └────────────┬─────────────┘
                              │
                              ▼
                      ┌────────────┐
                      │ API Gateway│
                      └──────┬─────┘
                             │
               ┌─────────────┼───────────────────┐
               │             │                   │
               ▼             ▼                   ▼
      ┌────────────────┐  ┌────────────────┐  ┌────────────────┐
      │ Upload Handler │  │ OCR Lambda     │  │ Q/A Lambda     │
      └───────┬────────┘  └───────┬────────┘  └───────┬────────┘
              │                   │                   │
              ▼                   ▼                   ▼
        ┌───────────┐     ┌──────────────┐     ┌──────────────┐
        │ S3 Upload │     │ Textract OCR │     │ LLM Summary   │
        └─────┬─────┘     └─────┬────────┘     └─────┬────────┘
              │                   │                   │
              ▼                   ▼                   ▼
        ┌──────────────┐   ┌──────────────┐   ┌────────────────┐
        │ Raw Text S3  │   │ DynamoDB Meta│   │ Bedrock/SageMaker│
        └──────────────┘   └──────────────┘   └────────────────┘

---

📁 Project Structure

smart-document-analyzer/
│
├── backend/
│   ├── textract-handler.py         # OCR + text extraction Lambda
│   ├── nlp-processor.py            # Summarization + embeddings
│   ├── query-handler.py            # Q/A Lambda function
│   └── utils/                      # Helper scripts
│
├── frontend/
│   ├── src/                        # React frontend
│   ├── public/
│   └── package.json
│
├── infrastructure/
│   ├── cdk/ or cloudformation/     # Infra as code
│   └── iam-policies/
│
├── docs/
│   ├── architecture.png
│   ├── demo-screenshots/
│   └── samples/
│
├── README.md
└── LICENSE

---

⚙️ Installation & Setup

1️⃣ Clone Repository

git clone https://github.com/<your-username>/smart-document-analyzer.git
cd smart-document-analyzer

2️⃣ Install Backend Dependencies

pip install -r backend/requirements.txt

3️⃣ Configure AWS Credentials

aws configure

Make sure your IAM user has permissions for:

• S3
• Lambda
• Textract
• DynamoDB
• Bedrock / SageMaker
• API Gateway

4️⃣ Deploy Backend Infrastructure

Using AWS CDK:

cd infrastructure/cdk
cdk deploy

5️⃣ Start Frontend

cd frontend
npm install
npm start

---

🧪 How the System Works

1. Upload Document

User uploads a PDF/image → sent to S3 via pre-signed URL.

2. OCR Trigger

S3 event triggers a Lambda function:

• Calls Amazon Textract
• Extracts text
• Saves cleaned text to S3

3. NLP Processing

A second Lambda:

• Chunks text
• Generates embeddings
• Creates summary using LLM
• Saves metadata to DynamoDB

4. Q&A Pipeline

User provides a question:

• System retrieves relevant text chunks (RAG)
• LLM generates the best answer

---

🧩 API Endpoints

Method	Endpoint	Description
POST	/upload	Generates pre-signed S3 upload URL
GET	/status/{docId}	Returns processing status + summary
POST	/ask/{docId}	Answers questions about the document

---

💰 AWS Cost Optimization

To stay within student credits ($199.78):

• Use Textract on small PDFs (≤ 5 pages)
• Delete SageMaker endpoints when not in use
• Prefer Bedrock for serverless LLM inference
• Enable S3 lifecycle rules to auto-delete temporary files
• Enable Billing alerts

---

📸 Screenshots (Add After Deployment)

Place images under docs/demo-screenshots/ and include examples here:

![Upload Page](docs/demo-screenshots/upload.png)
![Summary Example](docs/demo-screenshots/summary.png)
![Q&A Interface](docs/demo-screenshots/qa.png)

---

🧰 Tech Stack

Frontend

• React
• AWS Amplify

Backend

• AWS Lambda
• Amazon API Gateway
• Amazon S3
• Amazon DynamoDB
• Amazon Textract

AI

• Amazon Bedrock
  or
• AWS SageMaker JumpStart

---

🤝 Contributing

Contributions, issues, and feature requests are welcome!
Feel free to open a PR or fork the project.

---

📄 License

This project is licensed under the MIT License.
See the LICENSE file for full details.

---

👤 Author

SriSaiKiran Tambalkar
B.Tech CSE (AIML) Student
GitHub: https://github.com/
LinkedIn: https://www.linkedin.com/in