Cloud Computing at NASA's Frontier Development Lab

Abstract

NASA's Frontier Development Lab (FDL) is a research accelerator supported by NASA, the SETI Institute and industry partners. Each summer, FDL brings together teams of domain experts and machine learning scientists / engineers to work intensively for eight weeks to tackle some of the biggest challenges in space science, space exploration, and planetary protection. FDL solutions often require the training and deployment of deep neural networks, which are typically carried out on commercially available cloud compute infrastructure contributed by industry partners such as Google Cloud, Intel, IBM and NVIDIA. While FDL teams are co-located during the summer, collaborations persist for many more months, resulting in refereed journal, conference, and workshop publications and/or presentations. In this talk, the mentors of teams at NASA FDL and FDL Europe* will present case studies of how FDL teams use cloud storage and compute technologies for data preparation, rapid prototyping, and for scaling scientific and machine learning workflows to hundreds and thousands of machines . We also discuss how FDL teams use online tools (e.g., GitLab, Slack, Google Docs, Dropbox Papers) to facilitate effective remote collaboration. The domain areas covered in our case studies include astrobiology, exoplanet detection, space weather, lunar exploration and astronaut health monitoring.