modin-project/modin ? reverse-engineered prompt

Build me a Python library that acts like a drop in replacement for pandas, so I can change one import and make my existing dataframe code run faster on big datasets.

I want it to feel familiar to anyone who already uses pandas, with the same kind of dataframe API for common things like reading data, filtering, grouping, sorting, joining, and saving results. It should automatically use the machine’s available cores, and if possible let me choose between different compute backends like Ray, Dask, or MPI style execution. If a dataset is too large for memory, it should still be able to work sensibly instead of just failing.

Please include a simple quickstart, a few example notebooks or scripts, and clear install instructions for regular pip and conda users. Also add tests and basic docs so it’s obvious how to use it and how to switch from pandas with minimal code changes. If you need to check current backend docs online while implementing, go ahead and do that.