WTF (Who to Follow) is Twitter's user recommendation service, which is responsible for creating millions of connections daily between users based on shared interests, common connections, and other related factors. In this talk I will discuss the evolution of the WTF service: the first generation architecture depended on a system called Cassovary, an open-source in-memory graph processing engine built from scratch by Twitter specifically for WTF. This approach gave way to a Hadoop-based machine learning framework, which has recently been supplemented by a custom architecture for generating real-time recommendations. I will discuss the tradeoffs between different architectures, provide a general overview of algorithms, and share lessons learned in running a large-scale production service.