DocAI Platform

AI-Powered Document Intelligence API - A production-ready SaaS platform for document processing, OCR, and data extraction.

Live Demo | API Docs | Pricing

What is DocAI?

DocAI is a B2B SaaS product that enables businesses to extract structured data from documents using AI. Upload invoices, receipts, contracts, or forms - get back structured JSON.

Use Cases

• Invoice Processing - Extract vendor, amounts, line items, dates
• Receipt Digitization - Capture expenses automatically
• Contract Analysis - Pull key terms, dates, parties
• Form Processing - Digitize any structured form

Architecture

Edit in draw.io

Data Flow

1. Customer sends document via HTTPS to CloudFront CDN
2. API Gateway authenticates request via Cognito JWT
3. Upload Handler generates presigned URL, stores metadata in DynamoDB
4. Customer uploads document directly to S3
5. S3 Event triggers Document Processor Lambda
6. Textract performs OCR and data extraction
7. Results stored in DynamoDB, event sent to EventBridge
8. Webhook delivers results to customer's callback URL

Features

For Customers

Feature	Description
REST API	Simple JSON API for document upload and retrieval
Async Processing	Upload documents, get webhook when ready
Multi-format Support	PDF, PNG, JPG, TIFF
Confidence Scores	AI confidence for each extracted field
Custom Templates	Define extraction schemas for your documents
SDKs	Python, Node.js, Go client libraries

Platform Capabilities

Capability	Implementation
Multi-tenancy	Cognito user pools + DynamoDB tenant isolation
Authentication	JWT tokens via AWS Cognito
Rate Limiting	API Gateway throttling per pricing tier
Usage Tracking	DynamoDB counters + CloudWatch metrics
Webhooks	EventBridge → SNS/HTTP for async notifications
High Availability	Multi-AZ Lambda, DynamoDB on-demand

Pricing Tiers (Example)

Tier	Pages/Month	Rate Limit	Price
Free	100	10 req/min	$0
Starter	1,000	60 req/min	$49/mo
Business	10,000	300 req/min	$299/mo
Enterprise	Unlimited	Custom	Contact us

API Example

Upload Document

curl -X POST https://api.docai.example.com/v1/documents \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: multipart/form-data" \
  -F "file=@invoice.pdf" \
  -F "document_type=invoice"

Response

{
  "document_id": "doc_abc123",
  "status": "processing",
  "webhook_url": "https://your-app.com/webhook",
  "estimated_completion": "2024-01-15T10:30:00Z"
}

Extraction Result (via webhook or polling)

{
  "document_id": "doc_abc123",
  "status": "completed",
  "document_type": "invoice",
  "extracted_data": {
    "vendor": {
      "value": "Acme Corp",
      "confidence": 0.98
    },
    "invoice_number": {
      "value": "INV-2024-001",
      "confidence": 0.99
    },
    "total_amount": {
      "value": 1250.00,
      "currency": "USD",
      "confidence": 0.97
    },
    "line_items": [
      {
        "description": "Widget Pro",
        "quantity": 5,
        "unit_price": 250.00,
        "confidence": 0.95
      }
    ]
  },
  "processing_time_ms": 2340
}

Project Structure

docai-platform/
├── terraform/
│   ├── modules/
│   │   ├── networking/      # VPC configuration
│   │   ├── cognito/         # User authentication
│   │   ├── api-gateway/     # REST API with throttling
│   │   ├── lambda/          # Processing functions
│   │   ├── dynamodb/        # Data storage
│   │   ├── s3/              # Document storage
│   │   └── monitoring/      # CloudWatch dashboards
│   └── environments/
│       ├── dev/
│       └── prod/
├── src/
│   ├── api/                 # API handlers
│   ├── document-processor/  # AI extraction logic
│   └── webhooks/            # Notification handlers
├── docs/
│   ├── api-reference.md
│   └── integration-guide.md
└── .github/workflows/       # CI/CD

Technical Decisions

Decision	Choice	Why
Compute	Lambda	Pay-per-use, scales to zero, no ops overhead
Database	DynamoDB	Serverless, scales automatically, pay-per-request
Auth	Cognito	Managed user pools, JWT, integrates with API Gateway
OCR/AI	Textract	Managed, no ML ops, high accuracy
IaC	Terraform	Reproducible, multi-environment, team-friendly
Storage	S3	Cheap, durable, lifecycle policies

Deployment

# Deploy to dev
cd terraform/environments/dev
terraform init
terraform apply

# Deploy to prod
cd terraform/environments/prod
terraform init
terraform apply

Monitoring

Key Metrics

• Request latency (p50, p95, p99)
• Error rate by endpoint
• Documents processed per tenant
• Textract API costs
• Monthly active users

Alerts

• Error rate > 1%
• Latency p99 > 5s
• Cost anomaly detection
• Failed webhook deliveries

Security

• All data encrypted at rest (S3, DynamoDB)
• TLS 1.2+ for all API traffic
• Tenant data isolation at database level
• API keys rotatable per customer
• SOC 2 Type II compliant architecture

License

MIT

---

Built by: Venkata Subramanian Stack: AWS Lambda, API Gateway, DynamoDB, S3, Textract, Cognito, Terraform