Uber’s open source data visualization tool now goes beyond maps
You may not be aware, but Uber offers an open source version of the data visualization framework it uses internally, called deck.gl. The tool was made available to anyone via open source license last November, and now it’s getting some key updates that should help make it more useful to external teams and individuals looking for interesting ways to take their data and turn it into compelling visual representations.
“The main idea behind this library is that it’s a WebGL-powered framework that is designed for exploring and visualizing data assets at scale,” explains Nicolas Garcia Belmonte, Uber’s head of data visualization, regarding why the tool exists to begin with. “There’s a lot of geospatial stuff that we do here, as well, as you can probably imagine from the core business, so we visualize a lot of data on maps.”
Uber has used the tool internally for various purposes, including visualizing the pick-up and drop-off experience, as a way to make material the insights it can gather from the considerable stock of data it gathers from its ride hailing business. Open sourcing the project, according to Belmonte, is a way to help others be able to quickly and easily visualize and explore their own data sets, including very large collections.
The updates to the framework available today make it possible to go beyond maps, to visual data sets from machine learning and other more abstract use cases, including network traffic and more. The team also focused on usability, making it easier than ever to work with for developers with better documentation, as well as new demonstration projects.
deck.gl is a rich, flexible tool that according to Belmonte and Uber, has no real peers in the open source community. I asked them whether Uber had ever considered making this a paid offering, even though it’s relatively distant from their core business, given its capabilities.
“Personally, I’m a very big advocate for open source, so I think that open source can help the business in many other ways, and those ways are immediately more valuable than someone paying for this software,” Belmonte said. “I would say this is more about trying to reach out to developers and seeing what they can come up with in creative ways.”
Already, Uber has provided examples of how it can be used to visualize the points collected from a 3D indoor scan in exacting detail, or how it can represent visually Partial Dependence Plots in machine learning applications.