We study the problem of automatically finding gang member profiles on Twitter. We outline a process to curate one of the largest sets of verifiable gang member profiles that has ever been studied. A review of these profiles establishes differences in the language, images, YouTube links, and emoji features gang members use compared to the rest of the Twitter population. We generate word embeddings that translate these features into a real vector format amenable for machine learning classification and use them to train a series of supervised classifiers. Our classifiers achieve promising F1 scores with low false positive rates.