Nokia and Databricks have completed a joint trial in which they tested a unified data platform for autonomous telecom networks. The project addresses one of the biggest obstacles to AI applications at telecom operators: fragmented data spread across dozens to hundreds of separate systems.
Telecom networks generate large amounts of operational and business data. This information is often stored in various OSS and BSS systems, each with its own architecture and data sources. This makes it difficult to deploy AI models and automated decision-making on a large scale.
During the proof of concept, Nokia and Databricks investigated whether a common data architecture could function regardless of the underlying infrastructure. The premise was that operators should be able to use the same data streams and analysis tasks across different cloud environments or in their own data centers—without having to redevelop applications each time.
The technical trial focused on real-time performance management. It tested whether data pipelines could be deployed on different platforms without modifying the code. According to the companies, the same workflows could be executed on both Databricks and an open-source stack using Apache Flink, Kafka, and Iceberg.
In addition, Nokia developed an abstraction layer in Python. This keeps data transformations decoupled from specific platforms, allowing the same processes to be reused across multiple environments. An additional compiler then automatically translates the generic logic into platform-specific formats, such as Delta Live Tables or Flink SQL.
This approach is intended to prevent operators from becoming dependent on a single vendor or cloud platform. In the telecom sector, vendor lock-in has been a concern for years. This is especially true now that AI applications are becoming increasingly dependent on access to large amounts of network data.
AI agents as the next step
A second part of the trial focused on the use of AI agents to create new data products. According to the demonstration, an agent can automatically generate a data pipeline, have it verified, and then roll it out using natural language commands.
The companies view this as a potential building block for future autonomous networks, in which multiple AI agents collaborate. For example, an agent analyzing network issues could independently request or generate additional datasets to investigate the root cause.
Although the project is still in an experimental phase, it demonstrates the technical prerequisites needed to further integrate AI into telecom networks. The industry has long been exploring how to implement autonomous networks but often encounters limitations due to the lack of data interoperability between existing systems.
Nokia and Databricks intend to continue their collaboration and further investigate how AI applications can access real-time network data for analysis and automated decision-making. For now, however, this remains a technical validation; no concrete implementations by operators have been announced.
Also read: The Linux Foundation and Databricks launch OpenSharing