Micro1 Interview Question

How do you avoid leakage in preprocessing pipelines with sklearn? how do you ensure determinism in a multi-distributed GPU setting?