A while ago, I stumbled across a talk in an archive of a developer conference introducing a thin wrapper around the Numpy library that makes it more intuitive and "pythonic", especially when it comes to array operations involving multiple dimensions, appending/concatenating and axis arguments to various functions.
Unfortunately, I cannot remember the name of the library or the conference - it may have been FOSDEM, but I was unable to find it via a search there or on the internet.
I have seen JAX which does have slightly different/better indexing semantics than plain Numpy but is too rich/heavy for my needs.
