How to handle two projects that share the same pile of code

We have two repositories: one is a deployed product, and the other is some experimental ML project.

These two repositories are (and should) having the same preprocessing steps, say it is process.py. Synchronizing this module in two projects are hard to maintain because:

the same code is written in two different places, which is unnecessary
it's very open to mistakes because of #1
we have to update the production too if we change something in our experimental line.

What is the best practice to handle such a situation?

Do we have to create a third repo and keep our utility modules there and import it into these projects? Maybe a company-level repo that serves all of our utilities?
Should we create a testing repo identical to the production?
What are the other options?

Thanks.

My Stack