I am currently finishing a similar project, producing PDF from original Markdown input, with the added difficulty that input is spread over many files and directories. I chose LaTeX as the typesetting engine, since I know it fairly well and learning a completely different engine (Typst) with no clear advantage made no sense. Pandoc did a fine job with the low-level conversion from Markdown to LaTeX; then other files combined the generated .tex files into a coherent complete work.

Actually, the task was more difficult, because the input should also be used to produce a website -- and mkdocs was used for this.

I have a list of problems that came up and required not-insignificant amount of work to be solved. Many of them were about the handling of images.