While building our ML feature store we learned that a mix of databases yields significant gains in efficiency and operational simplicity.
Figuring out how to balance our experimentation speed with the necessary controls to maintain trust is never easy. Learn DoorDash's approach.
Learn how DoorDash developed a switchback testing method to perform incrementality testing on an app marketplace ad platform
Data preparation, represents The vast majority of work in developing machine learning models, learn how to make things easier
Learn how DoorDash build a platform to process billions of events from different data sources, quickly, consistently and reliably
Learn how DoorDash crowdsources data from a variety of sources to help predict realtime inventory for our new connivence and grocery product
When A/B testing is not recommended we can still quickly implement a new feature and measure its effects in a data-driven way.
Learn how DoorDash was able to test uniformly according to established best practices and reuse complex statistical methods with Dash AB
Learn about which caching libraries we considered, the analysis of our system and how we were able to use experiments to validate our approach.
Learn about the principles and big bets that enabled to scale and maintain our ML platform which supports our data users and data scientists
Ensuring ID's don't get exceeded is an evolving challenge. Learn how DoorDash made our tables were compatible with our new Bigint upgrade
Learn how our Fabricator infrastructure integrations for ML features were automated and continuously deployed to generate 500 unique features and 100+ billion daily feature values
Susbscribe to the DoorDash engineering blog