Real-Time Intelligence with IBM Time Series Models on Confluent

Foundation models are revolutionizing how businesses harness value from unstructured data, but the next frontier lies in stream-native intelligence. IBM and Confluent have joined forces to deliver this transformative capability, with time series foundation models (TSFM) now available in Early Access on Confluent Cloud. These models run alongside the data streams they analyze, eliminating the outdated approach of building and maintaining separate models for each domain.

Traditional decision-making often relies on outdated economics, with safety margins added to account for unforeseen events. TSFM changes this by generalizing to new time series data, enabling real-time forecasting, anomaly detection, optimization, and semantic intelligence. This accessibility means demand planners, fraud analysts, and process engineers can leverage these models without the need for large data science teams.

IBM has already demonstrated the effectiveness of these models in various industries, running them in its own products and operations before making them available to customers. The results are impressive: each point of accuracy translates to millions in savings, with productivity gains ranging from 5 to 10 times. Moreover, with domain experts handling decisions, the need for specialized data science teams diminishes.

The partnership between IBM and Confluent is built on a shared foundation: Confluent provides the real-time streaming platform, while IBM supplies the cutting-edge time series models. These models are designed to work seamlessly with Confluent's Flink SQL functions, enabling rapid deployment and integration into existing workflows.

The platform's native inference capabilities ensure secure, flexible, and cost-efficient real-time data processing, allowing teams to focus on developing ML applications rather than managing data infrastructure.