Powering PlayStation personalization to millions
summary
PlayStation consolidates scale by federating owner‑controlled lakes into a data ocean cataloged with AWS Lake Formation and Alation, ingesting 100+ billion events per day for 103 million active users. A lightweight ML platform trains hundreds of features per user, pushes attributes to a runtime feature store implemented on Aerospike, handling several terabytes for 100+ million users, millions of requests/sec with <10 ms latency; operationally the Aerospike clusters prove compact, low‑TCO and self‑healing while enabling recommendations, matchmaking and fraud detection.
AI-written summary · about
chapters (8)
- 0:00Intro
- 2:49PLAYSTATION TODAY
- 4:35FACING DATA CHALLENGES
- 7:17MACHINE LEARNING OPPORTUNITIES
- 8:15PLAYSTATION PERSONALIZATION
- 9:02FEATURE STORE REQUIREMENTS
- 10:00RICHT STORE TECHNOLOGY - CAP THEOREM
- 11:51SELECTING THE RIGHT STORE TECHNOLOGY
related talks
24mLarge Scale On-Demand Low-Latency Near Real-Time Predictions
42mBuilding a near real time personalized feed using data streaming and Aerospike
08mBuilding a Real-time Data Future
40mPowering Spotify's Audio Personalization Platform
21mSony Interactive Entertainment: Generative and predictive AI on AWS
29mUse sports data to elevate fan engagement and performance
15mFeature Stores: Powering Machine Learning Infrastructure with Apache Flink
26mPersonalisation at scale with AI/ML and Amazon Personalize