Low-Latency Recommendation Serving With ANN Retrieval and Ranking

System Design · Disney · Medium

Design a system that serves a recommendation model online, answering live recommendation requests. Traffic on the request path is dominated by reads, and every request must be answered within a tight latency budget. Because of this, the design must draw a clear line between the costly computation that can be done offline ahead of time and the work that has to happen for each individual request.

Loading the rest of the question…