By Jonah Harris
As a social discovery network with millions of daily active users, performing personalized recommendations at scale is a challenge. Unlike traditional one-way item-based recommendation systems, person-to-person recommendation is both reciprocal and behavior-based. Likewise, when you only have a user’s attention in the application for a few seconds, the ability to make split-second decisions based on real-time contextual information is essential.
This session consists of background regarding the challenges we’ve faced, the technology stack we’ve used to tackle such challenges, a detailed architecture of two Redis-based implementations, and the lessons we’ve learned.