Clinical research studying critical illness phenotypes relies on the identification of clinical syndromes defined by consensus definitions. A prime example is the pneumonia phenotype. Historically, identifying pneumonia has required manual chart review, a time and resource intensive process. The overall research goal is to develop automated approaches that accurately identify critical illness phenotypes. In this poster, we outline our approach to the identification of pneumonia from EMR, present our preliminary results, and describe future steps.