What is your issue?

No direct way to create multi-index dataset

Unless I'm missing something, there is no easy way to convert a Pandas dataframe to a dataset with a multi-index. For example, xr.Dataset.from_dataframe automatically converts any Pandas multi-index into separate dimensions.

Workaround is inefficient

One workaround is doing xr.Dataset.from_dataframe(df).stack(...) however this is very inefficient for sparse multi-indices.