Generative AI is revolutionizing the technology stack, introducing a new era of innovation and automation. Apache Flink is keeping up with this revolution through FLIP-437, which introduces AI models as first-class citizens in Flink SQL, on par with Tables, seamlessly integrating them into the existing resource hierarchy (Catalog->Database->Model). This groundbreaking advancement paves the way for unified data processing and model inference using the familiar SQL interface and positions Flink as a central piece of modern AI architectures.

In this 30-minute session, tailored for developers with experience in stream processing, we'll start by exploring the rationale behind Flink's decision to treat AI models as first-class entities within Flink SQL. We’ll then delve into the newly introduced ML_PREDICT and ML_EVALUATE functions, which enable a generic way for model inference and evaluation using Flink SQL. Finally, we’ll demonstrate how to leverage the powerful combination of Flink and Kafka to create an anomaly detection app that harnesses real-time data streams.

By the end of the session, attendees will understand how they can write an end-to-end data app with seamless AI inference integration using **just** Flink SQL.