IBM and Confluent make time-series models native to the platform
🎧 Listen to this story
IBM and Confluent have made a family of time-series foundation models natively available in Confluent for forecasting, anomaly detection and optimization. IBM says the models have surpassed 50 million downloads. They can be used in Confluent Cloud as a managed service or on premises with Confluent Platform for Apache Flink, and run on CPUs without requiring GPUs.
With Apache Flink, teams can prepare data and invoke the models using SQL, making the results available alongside the events that produced them. The approach targets use cases including keeping production in spec, detecting unusual store activity, product planning, assessing card payments and forecasting trading volume. Confluent’s Streamhouse architecture combines these models with continuously streamed data and other AI models to provide current context to applications, analytics and agents.
Why it matters · editorial interpretation
The integration lets teams invoke time-series models through SQL within Apache Flink workflows, either through a managed cloud service or on premises and without GPUs. This could simplify adding forecasting and anomaly detection to applications that need to work with continuously updated data.