IBM Time Series Foundation Models Now Available in Early Access on Confluent Cloud

Traditionally, analyzing time series data required experts to build individual models, a process that consumed significant time. In contrast, time series foundation models are trained on a vast array of signals, providing capabilities such as forecasting, anomaly detection, and optimization even for unseen data.

In this collaboration, the models are hosted on Confluent Cloud and invoked via Apache Flink. As a result, functions like demand forecasting and anomaly detection are provided as callable features rather than as projects that must be built from scratch.

IBM has been promoting the use of these models with partner companies in sectors such as cement, steel, food, and telecommunications. According to the company, productivity has increased by 5 to 10 times, enabling domain experts to make decisions without having to wait for specialized data scientists.

Currently, the models are accessible on Confluent Cloud via AWS, with plans to expand availability to Confluent Platform for on-premises and hybrid environments in the future.