I've heard PDFs aren't great in automated resume filters because the OCR doesn't work as well. Is it possible that you prefer PDF because of a survivor bias (PDFs that get past the filter are better for you but fewer PDFs get past it)?
I’m the founder of a startup that has eliminated the issue you mention. Our documents-to-database service handles arbitrary rotation, skews, and offsets.
Example of it handling a scan of a document that’s rotated ~100 degrees and physically cut in half with scissors here: https://siftrics.com/hydra.html
