One key method for producing nuanced interpretations of qualitative data, affinity diagramming, is ideal for wide-ranging, unstructured data. In this method, raw data, such as responses to interview questions and observations, are iteratively grouped from the bottom up, forming a hierarchy that facilitates understanding the data. An example affinity diagram is shown above, where a subset of student responses from a classroom exercise are grouped by similarity and labeled by statements that summarize the data. These labels are iteratively grouped until the top-level reveals high-level themes about the data. This process is relatively fast compared to methods such as qualitative coding. However, for large data sets, the diagramming process is inefficient and overwhelming. While a smaller diagram of 500 datum may take a few hours to build, larger diagrams (on the scale of thousands of datum) can take multiple days. Due to this limitation, researchers may avoid affinity diagramming if their data set is large, causing them to miss out on the nuanced data interpretations affinity diagramming can provide. Our team, in collaboration with Dr. Lydia Chilton, is working to develop a system called Qualitative Affinity Diagrammer (QuAD), which leverages natural language processing to generate suggested data groupings to address the scalability and efficiency of the diagramming process. This system is the first of its kind to harness deep learning to provide AI-powered recommendations for clustering qualitative data for affinity diagramming.