The Confluent Data Streaming Platform, particularly with Apache Flink for stream processing, offers a powerful tool for on-demand predictions in real-time applications such as fraud detection, customer personalization, predictive maintenance, and customer support. Remote model inference enables developers to connect real-time data streams to external machine learning models hosted on dedicated model servers and accessed via APIs, allowing for centralized model operations, streamlined updates, version control, and monitoring. This approach is ideal for high-throughput applications where latency is a trade-off for flexibility. Confluent's Flink AI model inference integrates remote AI models into data pipelines by calling AI models within Flink queries, managing remotely hosted AI models with SQL DDL statements, and invoking remote endpoints from various cloud services. The benefits of using data streaming for model inference include centralized model management, scalability and flexibility, efficient resource allocation, seamless monitoring and optimization, and reduced latency. This approach is particularly useful in hybrid cloud setups, where models might be hosted on a cloud-based infrastructure and accessed by edge or on-premises Flink applications. By separating the model server from the streaming application, developers can leverage powerful AI capabilities while keeping Flink applications focused on efficient data processing.