Although Transformer models could enable ML teams and other developers to build applications with unprecedented capabilities, ML practitioners lack tools for working with the unstructured datasets upon which Transformers are pretrained and fine-tuned. To fill this gap, Jina AI has released DocArray, a library that enables users to process, embed, search, store, and transfer nested, unstructured data (e.g. text, image, audio, video) with a Python API. DocArray relies upon two concepts: Documents (data structure to represent nested, unstructured data) and DocumentArrays (container to wrangle multiple documents).