At Uber, we take advanced research work and use it to solve real world problems. In our Science at Uber video series, Uber employees talk about how we apply data science, artificial intelligence, machine learning, and other innovative technologies in our daily work.
Machine learning helps Uber make data-driven decisions which not only enable services such as ridesharing, but also financial planning and other core business needs. Our machine learning platform, Michelangelo, lets teams across the company train, evaluate, and deploy models that help us forecast a wide range of business metrics.
As a platform, Michelangelo’s management of data pipelines and workflows helps teams make our models as accurate as possible. The platform enables our teams to simply, flexibly, and intelligently prototype and productionize machine learning solutions at scale with tools such as Horovod, PyML, and Manifold.
Product manager Logan Jeya knows the ins and outs of Michelangelo, having worked on the project from when it only hosted a handful of models to its current scale of thousands of models. Now, the team behind Michelangelo has grown to over 40 engineers and its models can train using over a billion records.
Check out our other Science at Uber videos:
- Science at Uber: Building a Data Science Platform at Uber
- Science at Uber: Bringing Research to the Roads
- Science at Uber: Powering Uber’s Ridesharing Technologies Through Mapping
- Science at Uber: Applying Artificial Intelligence at Uber
- Science at Uber: Making a Real-world Impact with Data Science
- Science at Uber: Innovating Across Digital and Physical Worlds