Machine Learning and Real-Time Analytics in Apache Kafka Applications

This post introduces two primary methods for deploying machine learning models within Apache Kafka applications: model servers with RPC and embedding models directly into Kafka client applications. It details the use of TensorFlow Serving for the model server approach, providing code snippets for Kafka Streams RPC. For embedding models, it demonstrates loading TensorFlow models directly into Kafka Streams applications and discusses KSQL UDFs for model deployment. The post also outlines criteria for deciding when to embed models (performance, size, binary format) and the trade-offs between the two approaches, including latency, availability, and management features. It concludes by mentioning cloud-native deployment strategies with Kubernetes.