(T) Uber hosted this month a meet-up to share its various internal machine learning initiatives on September 12 at its headquarters in San Francisco. Uber mission is to provide reliable transport for everyone and everywhere. Uber is present today in six hundred cities worldwide.
Following is a summary of the meet-up with the videos of the presentations.
Approach to ML at Uber
ML at Uber is extensively used to provide an efficient ride-sharing marketplace. Uber ecosystem is a space-time environment that needs to scale to billions of calculations, and thousands of decisions made for millions of riders and drivers every minute.
Designing Uber Maps
Maps are critical to every component of Uber transportation network: from destination search and ride prediction, generation of map tiles, ETAs, routing, and up-front fare estimates, maps are critical to every component of Uber transportation network:
Uber Marketplace is where riders meet their drivers in the Uber mobile app. Marketplace leverages many ML algorithms to meet the real-time supply and demand of the Uber transportation network, and in particular:
- Demand Modeling
- Dynamic Pricing
Uber Machine Learning Platforms
Uber has built different ML platforms for its internal applications and in particular:
- Michelangelo its ML-as-a-service platform designed to cover the end-to-end ML workflow: manage data, train, evaluate, and deploy models, make predictions, and monitor predictions. Michelangelo also supports traditional ML models, time series forecasting, and deep learning
- A Natural Language Processing (NLP) platform to generate and deploy actionable responses for its customer support tickets, chatbots to make driver onboarding easier, and suggested in-app replies
- Anomaly detection to ensure reliability of Uber services during all hours including day-night and weekday-weekend cycles
Reference: Uber Engineering Blog
Note: The picture above is from the meet-up.
Copyright © 2005-2017 by Serge-Paul Carrasco. All rights reserved.
Contact Us: asvinsider at gmail dot com.
Categories: Machine Learning