As science becomes increasingly data-driven, software plays an increasingly important role. However, a gap in the pipeline has emerged: faculty, students, and postdocs in many scientific domains are not equipped to develop and deliver the advanced software they require. Furthermore, even in computer science, conventional academic roles have little incentive to harden, sustain, share, and integrate their techniques into a robust, reusable software infrastructure.
The charge of this working group was to fill this gap, the goal being to sidestep inefficiencies to software development arising from competitive funding, including overemphasis on novelty, “not-invented-here” syndrome, underemphasis on usability, and a tendency towards overgeneralization or overspecialization.
Together with our counterparts at the University of Washington and New York University, this group conducted a distributed experiment varying people, policies, procedures, and projects to find foundational methods for the development, delivery, and sustainability of science software. Through shared events, mutual testing and outreach, shared software licensing and IP policies, and more, the group planned to develop high-impact, usable software in domain science; establish ourselves as national leaders in open source software excellence, and attract and retain top data science “builders.”
Original working group members: Stéfan van der Walt, David Ackerly, Carl Boettiger, Sandrine Dudoit, Fernando Pérez, Saul Perlmutter, Karthik Ram, Daniela Ushizima