I've been using "pdfcpu optimize" as a post-processing step for dynamically generated pdfs for several years now. Simple, well-documented, and fast. I've been very happy with it