NSF-PAR, among other, requires PDF/A documents. They prodive some indications on how to produce such documents, but unfortunately only using Microsoft Word Documents and Adobe Acrobat Professional.
If you can access a .tex source code, then you can use latex2pdfa.
If you can only access the .pdf (for example of the published version), then you can use OCRmyPDF, for example with
ocrmypdf --tesseract-timeout=0 --skip-text input.pdf output.pdf
so that your document is not "OCRized".