Airbnb utilizes its User Signals Platform (USP) to handle real-time personalization effectively, processing over 1 million events per second. USP ingests and processes user actions with sub-second latency, stores both real-time and historical data, and provides a configurable interface for developers. Key features include the use of Flink for event-driven processing, an append-only data model for resilience, and a robust infrastructure with hot standby Task Managers for operational stability.
Table of contents
Build Private AI Agents at Scale (Sponsored)Architecture OverviewKey Engineering DecisionsUser Signal TypesUser SegmentsSession EngagementsFlink Stability with Hot StandbyConclusionSPONSOR USSort: