Learn how our Revenue Platform team re-built Twitter’s Ad Platform Architecture for the future.

Twitter has recently built a streaming logging pipeline that facilitates the Home Timeline ranking model training. The new pipeline not only shortens the latency but also improves the resiliency.

Twitter’s traffic has had a dramatic increase over the course of 2020. South American traffic peaked at roughly 3x our available capacity. We describe how we survived.

We explore state-of-the-art metric learning ML techniques recently launched to production which improve ad relevance, helping Twitter better serve the right ad to the right user at the right time.

The challenges the Search Infrastructure team at Twitter went through in order to reduce the search indexing latency to one second.

Twitter discusses a simple machine learning model that prioritizes ad requests based on how much revenue we expect to make from them.

Twitter shares tips for deleting data in a microservices architecture using an erasure pipeline.

This outlines the whole debugging process of a Linux kernel bug we found during an incident and demonstrates how we worked with Linux kernel open source community.

As parts of the Twitter application grow, we can scale demands on capacity by adding more instances or replicas to a respective service cluster (i.e., horizontal scaling)

A design pattern that can be applied to all service types to increase team productivity.