We are using AWS Textract to OCR PDFs. We don’t use any of their advanced form/table detection, queries, etc. features. We only want to take a PDF (or PNG/JPEG), and get back a list of all “blocks” with coords that are numbers.