BoxKit is a library that provides building blocks to parallelize and scale data science, statistical analysis, and machine learning applications for block-structured simulation datasets. Spatial data from simulations can be accessed and managed using tools available in this library to interface with packages like SciKit, PyTorch, and OpticalFlow for post-processing and analysis. The library provides a Python interface to efficiently access Adaptive Mesh Refinement (AMR) data typical of simulation outputs, and leverages multiprocessing libraries like JobLib and Dask to scale analysis on Non-Uniform Memory Access (NUMA) and distributed computing architectures.