Feature Stores at Spotify: Building & Scaling a Centralized Platform

apply(conf) - Apr '21 - 10 minutes

Over 345 million Spotify users rely on Spotify’s great recommendations and personalized features in 170 different markets around the globe (with 85 of these markets launching in the first part of 2021). Some users even claim Spotify knows their tastes better than they know them themselves! How does Spotify build these great recommendations? Unsurprisingly, with data and machine learning! But with the massive inflows of data and complexity of production use cases, defining a unified approach to ML is challenging. In this talk, Aman will give an overview of the challenges we face with building a central ML Platform at a highly autonomous organization, and our approach of adoption by incentive. As an example, we will dive deeper into how this motivates a Feature Marketplace strategy at Spotify (our history with feature tooling, the foundation we are building now and where we are headed).

Aman Khan

Product Manager


Aman is a Product Manager on the ML Platform team at Spotify. Aman is focused on operationalizing ML feature management tooling to ensure that Spotify uses data for ML workflows effectively. Prior to Spotify, Aman was a Product Manager at Cruise where he built frameworks to test self-driving cars at scale. Aman has a BS in Mechanical Engineering from UC Berkeley, and a background in Product Design which he applies towards solving technical and product challenges in the ML space.