Unbiased all-sky surveys such as the Zwicky Transient Facility (ZTF) or the Pan-STARRS Survey for Transients (PSST) have opened up the door for the discovery of new and exciting types of transients. The current discovery rate of optical transients makes it such that only a small fraction of them can get spectroscopically classified, and by the time the Legacy Survey of Space and Time (LSST) commences, the number of discovered transients is expected to increase by about two orders of magnitude. We have been running a program to follow up alerts from these streams in search of superluminous supernovae (SLSNe), since only a handful of these are known, and many questions remain open regarding their power source, progenitors, and diversity of features in their light curves. In the process of searching for SLSNe we have also encountered other exotic transients, such as tidal disruption events (TDEs) and a pair-instability supernova candidate. In order to decide which transients are most worthy of follow-up, we have developed a custom machine-learning pipeline to estimate how likely any new transient is to be a SLSNe, and in this way make the most efficient use of our telescope resources.