Cluster convolutional neural networks (CCNNs) have many advantages over single, one-shot neural networks that may be capitalized on in edge devices. Some industries investigating the use of machine learning into their products entertain applications where the consequences of low sensitivity or specificity in a model may be catastrophic. In some of these scenarios, processing capability is minimal and there is no availability for cloud transactions. In these instances, conventional convolutional neural networks (CNNs) may not fit the bill. Here, we propose a machine learning model architecture that produces increased accuracy at lower computational cost than conventional CNNs. We establish how this architecture is inherently smaller in size by design without compromising accuracy, opening artificial intelligence models up to a much larger population of edge devices and solutions in medicine. We illustrate our proof of concept of this architecture through the identification of seven respiratory conditions in chest X-ray images by utilizing the machine learning model in a mobile app. Additionally, we elaborate on how the model better protects data privacy, may be trained on much less data than typical neural networks, and allows for explainability of results. Finally, we evaluate the efficacy our model through a clinical trial and compare its results to a conventional CNN on the same data.