You know how user profiles contain all the info you need from day 1, including future attributes you never envisioned needing?

Neither do I.

User profiles are a crucial building block for any personalized recommendation algorithm. Especially when you're the scale of Outbrain, with over 1 billion named users. But the brilliant architecture that we had for years evolved into a brilliant, steaming mess that slowed us down to a screeching halt. That is, until we bit the bullet and redesigned this core component in our architecture stack.

In this talk we'll show you how to create AB-tesable profiles, filter signal from noise and prevent data from turning you into a cat lady.