Student Learning Hub: Introduction to Online Recommendation Systems: From Predict-Sort Architecture to Real-Time Feature Service

This session introduces how modern online recommendation systems, like TikTok shop, deliver personalized content in real time. I’ll walk through the overall system architecture and then focus on the predict-sort stage, explaining how BFS (Bytedance Feature Service) and UDA (User Data Accessor) enable large-scale feature computation using a DAG-based (Directed Acyclic Graph) operator framework and a domain-specific language (DSL). Participants will gain insight into how these components work together to achieve millisecond-level predictions. Hosted by NYU’s Hongdao Meng!