Advances in telescopes, detectors, and computing processing power has led to a new field of astronomical study called time-domain astronomy. Time domain means that we continuously monitor large regions of the sky in order to identify dynamic events: explosions, flares, rotations, or orbiting sources. In the coming era, the discovery potential is growing rapidly with new surveys such as PTF2, Catalina Sky Survey, PanSTARRS, and LSST -- yet follow-up capability is not growing at the same pace. It turns out that the best way to leverage precious follow-up capability is to screen objects to determine if they are interesting or not. In this talk I describe the SED Machine, a spectrograph for classifying transients.