Strip metadata and re-save with optimised object streams. Lossless — no quality change to text or images.
Drops your PDF's metadata (title, author, subject, keywords, producer) and re-saves with object-stream compression on. Savings are typically 5–30% on PDFs that were exported from Word or browser "Print to PDF" without optimisation. Image-heavy PDFs benefit less because we can't re-encode their JPGs in pure JavaScript without quality loss.
For aggressive image re-compression, use a desktop tool with Ghostscript or upload to a server-side compressor. This in-browser version trades some compression power for the privacy guarantee that your file never leaves your device.