From raw data to informed intelligence: Democratizing data science and ML at Uber

Description

Uber is changing the way people think about transportation. As an integral part of the logistical fabric in 600+ cities across 65 countries around the world, its using technology to give people what they want, when they want it. So whether it’s a ride, a sandwich, or a package, its systems are tirelessly optimizing every part of the journey to ensure each experience is nothing short of magical.

To do this, teams across Uber depend on data to power every insight and inform every decision. People working with data at Uber bring various skills and proficiencies to the table: some advanced users know exactly what they’re looking for, while others are learning to explore various techniques to wrangle and make sense of data. But they all have one thing in common: they want to make intelligent data-driven decisions. And Uber has been at the leading edge of democratizing data science and seamlessly productizing machine learning since 2015. Every aspect of the Uber experience is powered by data and ML—everything from in-app ETAs, menu recommendations, and map labeling to driver dispatch and customer support.

Atul Gupte explores how Uber has approached building data science tools by working with its most advanced users to democratize complex techniques for less-technical people. He then dives into why Uber believes it’s important to offer seamless pathways to productionizing ML models with minimal effort so teams can leverage this to power various experiences. And you’ll hear several impactful examples. Atul dives into a few key items in particular, including Python model serving at scale (PyML), hyperparameter optimization (autotune), deep learning in the cloud (Google Cloud Platform [GCP]), knowledge feed of notebooks and dashboards, the challenges of reliability and monitoring at Uber’s scale, and concrete use cases in production and new ones coming down the pike.