Inference or production mode for trained model

Dear all,
Good day.
I wish to clarify whether my understanding regarding the workflow is correct or not.
For example, I implemented MCMC and get the trace.

I saved the trained model to_netcdf.

In a new file, I import the saved_netcdf using from_netcdf.
I will fit a model and some new inputs, perform sample_posterior_predictive to get new predictions
Is it correct?
Please correct me if my understanding is wrong.
Please share the experience or tips on how to implement the trained model in proper way.

Thank you very much