XND is an extensible framework of C libraries and Python modules for

handling typed memory blocks. XND recreates the foundations of NumPy as a

number of smaller libraries, combining the lessons learned in the past twenty

years of array computing in Python with the needs of newer applications.

XND's type system is more general than NumPy's and allows unpacking

arbitrarily nested Python type hierarchies directly to memory.

XND's computational kernels are multimethods and can be generated

and inserted into modular lookup tables at runtime. This opens the

possibility for a large variety of applications written in different

languages to use XND as a container backend.

Ongoing work connects

XND to Numba, Lapack/MKL and the CUDA kernel libraries. There is

interest in integrating XND with a variety of libraries like Dask,

xarray, Numba, Chainer, PyTorch, Tensorflow, PyMC4, TVM/NNVM,

Plasma Store, Apache Arrow, and Tensor Comprehensions.