CyberTraining: CIU: The LSST Data Science Fellowship Program

We propose an innovative education program that bridges the gap between traditional astronomy graduate curricula and the essential computing skills of the modern astronomer. This program addresses a critical need in the astronomical community, and will enable better use of and scientific output from the NSF’s flagship astronomical observatory for the next decade, the Large Synoptic Survey Telescope (LSST). LSST is poised to bring a data revolution to astronomy, not just by creating so-called “big data”, but in the sense of transforming the best practices for research in the field. Currently, most astronomers are self-taught with respect to computing, resulting in ad hoc practices for code distribution and management, and little thought (except in isolated subfields) to efficient algorithm design, implementing proper statistical frameworks, or managing and maintaining data access. While there is awareness of LSST and a desire to improve these skills amongst current graduate students, graduate curricula have been slow to evolve. The LSST Data Science Fellowship program is a two year training program designed to teach the skills required for LSST science that are not easily addressed by current astrophysics programs. The program consists of three, one-week schools per year over a two year period. Our curriculum covers the basics of managing and building code, statistics, machine learning, scalable programming, databases and data management, image processing, visualization, and communication; curriculum materials are distributed openly to better enable adoption by the field at large. Week-long sessions over multiple years allow our students to do deep dives on the course material, reinforce concepts from previous sessions, and put their learning into practice, in addition to building community around these skillsets, teaching peer mentoring, and planting the seeds of future collaborations. In the following proposal, we detail the program structure and motivation, and provide an initial analysis of results from our pilot program.