project-monai/monai ? reverse-engineered prompt

Build me a Python library for medical imaging AI that feels like a practical toolkit for researchers and engineers working with scans and other healthcare images. I want it centered around PyTorch, with clean, reusable pieces for loading and preprocessing multi dimensional medical images, building and training models, evaluating them with useful metrics, and plugging in common losses and network components without a lot of boilerplate.

It should support end to end training workflows, be flexible enough for both beginners and more advanced users, and work well on GPUs, including scaling to more than one GPU when needed. Please make it installable as a normal Python package, add solid docs and examples so someone can get started quickly, and include tests so the core features are reliable.

If it makes sense, also set up a Docker option for an easy run anywhere environment. Keep the code organized and production ready, but still friendly for research use. Look up current docs online if you need to.