Astronomy has been propelled to the frontier of data-intensive science by advances in computing as well as telescopes. With its 3,200-megapixel camera, the upcoming Large Synoptic Survey Telescope (LSST) will photograph 10 square degrees of sky every 17 seconds. Hidden in the resulting 15 TB-per-night stream of images are patterns in how the universe evolves. In order to make astrophysical measurements—such as the number density of supermassive black holes in the early universe—we first convert this movie of the sky into a database of flux measurements using image-processing algorithms (e.g. change detection, background subtraction, deblending). We will discuss statistical challenges in both processing noisy, heterogeneous images and classifying objects using ensemble classifiers on the resulting time-series datasets.