How Jumio built a real-time feature store on AWS
Sub-100ms feature serving at scale: how Jumio cut fraud-detection latency and saved $120K annually with a real-time feature store.
Why it matters
A concrete ML infrastructure case study showing how practitioners can architect real-time feature pipelines for production AI/ML workloads on AWS—actionable for teams building fraud detection, recommendation, or risk-scoring systems.
The key facts
8 to knowSub-100ms feature serving latency achieved
$120,000 annual cost savings
Tech stack: SageMaker Feature Store, Managed Flink, Kinesis Data Streams
Use case: fraud detection (real-time inference requirement)
AWS blog case study (vendor content but with technical depth)
Real-time feature store on AWS (SageMaker Feature Store + Flink + Kinesis)
Use case: fraud detection
Published as AWS blog case study
Go to the source
AWS Machine Learning Blogaws.amazon.com
Publisher excerpt: Learn how Jumio built a centralized, real-time feature store on AWS with Amazon SageMaker Feature Store, Amazon Managed Service for Apache Flink, and Amazon Kinesis Data Streams. The architecture delivers sub-100ms feature serving for fraud detection and saves approximately $120,000 annually.