
Build Less, Design More: Composable Visualization Systems
Dominik Moritz, faculty at Carnegie Mellon University and visualization team lead at Apple, argues that data visualization research should stop building monolithic, single-purpose systems and instead design shared specifications and APIs. Speaking in Stanford's CS547 Human-Computer Interaction Seminar, Moritz traces how composable frameworks built on declarative interfaces, like his own Vega-Lite, Falcon, Draco, and Voyager projects, let data scientists and machine learning engineers assemble ephemeral interfaces suited to the task at hand rather than waiting on bespoke tools. He makes the case that specifications, while intellectually demanding to design, produce more reproducible research and a healthier tool ecosystem than isolated systems built from scratch each time. Drawing on his background at the University of Washington under Jeff Heer and Bill Howe, and his current work spanning academia and industry, Moritz connects abstract interface design choices to concrete outcomes for working data scientists in Python and JavaScript communities.