It doesn't matter how much of an expert you are if your algorithm fundamentally isn't parallel though. We just don't know how to parallelise some things.

Fully agreed, but often these can be solved by reviewing the overall (business) problem.

E.g. maybe it's hard / not worth it to process a file in parallel, but maybe the most frequent use case needs to process more than one file and files can be processed in parallel with much less work.