Real-time machine learning with ScyllaDB

Meetup link: https://www.meetup.com/vienna-data-engineering-meetup/events/307457631/

Real-time ML systems rely on fast access to features. Feature stores provide the infrastructure to manage and serve these features, but latency, throughput, and workload diversity pose significant challenges. This talk explains what features are and why feature stores are essential for modern, large-scale ML pipelines. We will cover online and offline feature store workloads, the performance requirements of real-time inference, and the role of low-latency databases.

We’ll cover ScyllaDB, a high-performance NoSQL database, and how it supports online feature store use cases through integration with the Feast framework.