Contact person: AI OCR PDF Reporting API

Phone:Show

Email:Show

Location: Charlotte, United States

Budget: Recommended by industry experts

Time to start: As soon as possible

Project description:
"I need a clean, well-documented REST API that ingests PDF documents—primarily invoices, contracts, and receipts—runs them through reliable OCR, layers on AI post-processing, and returns a single Excel file that captures everything found in the originals.

Core workflow
1. The endpoint receives one or many PDFs.
2. OCR extracts all text, detects tables (with cell-accurate structure), and pulls out embedded images or graphics.
3. An AI layer normalises field names, fixes OCR errors, and tags each data block by type (line item, signature block, logo, etc.).
4. A neatly formatted .xlsx workbook is generated: one sheet per document with raw text, one sheet for each table in native Excel table format, and hyperlinks or thumbnails for extracted images. The final workbook is streamed back to the caller.

Technical considerations
• Language and stack are up to you, but please choose mature libraries—e.g., Tesseract, Google Vision API, AWS Textract, or similar for OCR, and any modern ML/NLP toolkit for the AI clean-up.
• The service must accept PDFs up to 20 MB, process at least three files in parallel, and respond with a presigned download URL.
• Return codes and error messages should follow standard HTTP semantics so integration is straightforward.

Deliverables
• Source code for the API (with Dockerfile).
• Brief setup guide plus endpoint documentation (OpenAPI / Swagger preferred).
• Sample Excel report generated from three supplied PDF examples.
• Short README explaining how the AI layer can be retrained or swapped if future needs change.

Acceptance criteria
• 95 % of text blocks must appear in the Excel output.
• Tables preserved with correct row/column counts.
• Images extracted and linked in the workbook.
• All three sample documents processed in under two minutes on a modest cloud instance.

If you have prior work combining OCR and AI for document intelligence, let me know; re-usable models or pipelines are welcome so long as licensing is clear." (client-provided description)