The organization of neural circuitry plays a role in brain function. Investigations into this organization have generally had to trade off between coarse descriptions at a large scale and fine descriptions on a very small scale, due to the constraints of the tools and data processing methods available. Recently, reconstructions of tens to hundreds of thousands of neurons at synaptic resolution have been published, enabling investigations into the interplay between global, modular organization, and cell type-specific wiring. Analysis of these enormous data sets requires the development of new methods and tools, to address the complexity and scale of the data. We present such a method, based on a machine learning algorithm developed by Guo et al. We have applied novel community detection methods to the Hemibrain data set, a synapse-level reconstruction of 20 thousand neurons and over 3.5 million synaptic connections in the brain of the fruit fly (Drosophila melanogaster). By optimizing a modularity score across a variety of resolutions, we are able to examine the hierarchical community structure of this network, revealing both large-scale (on the order of thousands of neurons) and small-scale (on the order of tens of neurons) structure in the network. We generally find a high degree of modularity throughout the network. Our methods are capable of identifying well-known features of the fly brain's anatomy and sensory pathways. For example, our method automatically identifies and refines the layered structure of the fan-shaped body in an unsupervised way. These methods show that the fine-scale, local network reconstruction made possible by modern experimental methods are sufficiently detailed to identify large-scale, global organization of the brain. They moreover enable automated, large-scale generation of novel predictions of brain organization.