Machine Learning Infrastructure at Twitter

Machine Learning (ML) is used at Twitter in many different ways. The problem domain is diverse - finding interesting Tweets for each user; preventing abuse; classifying media (photos). These tasks are performed by different types of Machine Learning models.

In this talk, I describe the challenges in creating a platform (set of software and services in the Cloud as well as inside Twitter Data Center) that helps Twitter Engineers use Machine Learning to solve a diverse set of problems.