In this blog post, we discuss the scale at which Twitter operates through Project Sparrow, an initiative that shifted the architecture of our data pipelines from a batch event approach to streaming.

Over-squashing is a common plight of Graph Neural Networks. In this post, we discuss how this phenomenon can be understood and remedied through the concept of Ricci curvature.

In this blog, we share findings from a follow-up analysis in which we seek to understand how prompts cause people to reconsider potentially harmful or offensive content before they hit send.

In this blog, we discuss how we approached migrating an exabyte of data to Google Cloud to make it easier for our Tweeps to analyze and visualize data.

In this blog, we share an overview of Twitter’s Druid ecosystem and discuss our work towards a unified ingestion experience.

On the Data Science team, a piece of our job is understanding how conversations are happening on Twitter. In this blog, we use Wordle as a case study to showcase how we think about similar analyses.

This blog post discusses how feature propagation can be an efficient and scalable approach for handling missing features in graph machine learning applications.

In this blog post, we talk about a performance and stability problem we encountered while migrating Manhattan's storage engine to RocksDB and how we solved it.

In this blog, we share details about how we built an in-house product called Qurious, which allows our internal customers to get answers to their analytical queries through natural language questions.

Twitter’s new Thrift library is an open-source, standalone, lightweight, data encoding library. In this blog, we share our library so iOS developers outside Twitter can start using Thrift data.