The Massive Open Online Course (MOOC) movement has catalyzed discussions of digital learning on campuses around the world and highlighted the increasingly large, complex datasets related to learning. This talk will highlight the MOOC movement from the perspective of a researcher studying learner behavior, a developer of big data pipelines capable of handling billions of data points, and above all, as a physicist working in the education field. The talk will begin with a MOOC research retrospective, reviewing our understanding of the nearly 7 million unique users that have led to over 13 million enrollments across more than 300 open online courses from Harvard and MIT. From there, I will share preliminary efforts to better inform instructors how design choices impact learning outcomes. By abstracting - or coarse-graining - MOOC content into simple linear sequences of course components (videos, problems, pages), we are able to cluster sequences and analyze variance in learning outcomes. Lastly, in order to further motivate the impact of big data in education, I will highlight new projects at Harvard aiming to transform current models of teaching and learning. My hope for this talk is that listeners will better understand the current wave of digital education and the opportunities it provides for data-driven teaching and learning.