LeanOps is a serverless, event-driven document processing pipeline designed to help small businesses eliminate manual data entry using AI.
It automatically converts unstructured PDFs (invoices, contracts, applications) into clean, structured data — with no servers to manage and near-zero operational overhead.
Small companies often waste 10+ hours per week manually entering data from PDFs into spreadsheets.
- Hiring human operators is expensive and does not scale
- Traditional OCR software breaks on varying document layouts
- Inconsistent vendor formats lead to errors and rework
LeanOps uses AWS serverless services and LLM reasoning to extract, understand, and structure data from messy PDFs automatically.
Key benefits:
- 💸 Cost: ~$0.02 per document
- ⚙️ Maintenance: Zero server management
- ⚡ Speed: From upload to database in < 10 seconds
- 🧠 Intelligence: Uses AI reasoning, not rigid templates
- AWS Lambda – Orchestrates the event-driven workflow
- Amazon Textract – Extracts raw text from complex PDF layouts
- Amazon Bedrock (Claude 3.5 Sonnet) – Parses text into clean, structured JSON
- Amazon DynamoDB – Stores extracted data for analytics and dashboards
- Amazon S3 – Entry point and trigger for the automation pipeline
- PDF uploaded to S3
- S3 event triggers Lambda
- Lambda calls Textract for text extraction
- Extracted text is passed to Bedrock (LLM) for semantic parsing
- Clean JSON is stored in DynamoDB
- Data becomes immediately available for downstream systems
git clone https://github.com/ramiro85/leanops-ai
cd leanops-ai
# Build the project (use-container ensures dependencies match Lambda's environment)
sam build --use-container
# Deploy the stack (first time use --guided)
sam deploy --guided