Building intelligence: a data platform strategy for autonomous networks

Executive summary:

This article presents a data platform strategy for Communication Service Providers (CSPs) transitioning toward autonomous networks. The article emphasizes the critical role of data platforms in enabling cross-domain orchestration and insights across Network domains. Although not the focus of this article, the same apply across I.T/Business & Customer experience domains and their cross domain insights. High level architecture and technical components are discussed to guide CSPs in their journey towards autonomous networks

Data lakes & data mesh, the autonomous networks backbone

With autonomous networks, the ground is shifting to data and intelligence.

Autonomous networks require a unified data fabric, a platform that enables seamless ingestion, processing, and insights across every network domain and layer.

To enable self-configuring, self-healing, and intent-driven networks, CSPs need to embrace a modern data architecture with data lakes and data mesh as the key foundational design patterns.

The data platform architecture

Autonomous networks rely on real-time telemetry, predictive analytics, and closed-loop automation. To achieve autonomy, operators need to architect networks as data-first ecosystems where data is the strategic asset. This requires two critical components:

1. A scalable data lake to centralize and unify massive volumes of raw telemetry, logs, and events.
2. A data mesh architecture to decentralize ownership, ensure data quality at source, and enable agility across domains.

A data lake acts as the central repository for all structured and unstructured network data—collected at high velocity and scale from RAN nodes, core networks, edge devices, and user interactions.

In the context of autonomous networks, the data lake becomes the training ground and decision-making layer for predictive maintenance and intent-based orchestration.

While data lakes offer centralization, data mesh architecture brings autonomy and accountability to data at the domain level. Instead of one monolithic data platform team, each network domain (E.g. RAN, Transport, Core, Edge) becomes a data product owner, responsible for producing and maintaining high-quality, discoverable, and reusable data sets.

Autonomous networks rely on localized intelligence such as real-time optimization at the edge, or fault prediction in core. Data mesh allows distributed ML pipelines close to the data sources, improving latency and performance.

Additionally, a self-serve data infrastructure consisting of Common infrastructure (e.g., streaming platforms, data lake connectors, lineage tools) enables teams to produce and consume data independently and Federated Governance forms unified standards for data privacy, policy enforcement, and compliance.

Federated data governance spans all domains with common standards (schemas, metadata, access control, quality, compliances, monitoring). All data products are registered in central catalog, tagged with domain, owner, description, freshness

With increasing autonomy and resulting increased exposure, the platform must enforce policy-driven data access, ensure data lineage, and maintain compliance with standards like GDPR and sector-specific regulations. Trust is a prerequisite for automation.

Domain oriented data products need to discoverable (catalog), addressable through APIs/data stream, governed, secure, need to have well defined contracts and meta data. In the case of data mesh standardized data Contracts are critical.

Depending on size of the CSP Operations and stage of the autonomous journey, smaller CSPs may find a unified data lake a better fit while larger CSPs may lean towards separate domain data lake with a data mesh approach. Long term, a separate data mesh is a better approach for scalability and domain ownership. However, even larger CSP can start off with a unified single data lake and later evolve.

Strategic nature of architecture

Combining a centralized data lake with a decentralized data mesh offers the best of both worlds for scalability, resilience and speed to insights

A new operating model and cultural shift for CSPs

CSPs must evolve from central data teams to a network of data product teams, with shared principles but localized autonomy.

This requires:

• New roles (e.g., data product owner for RAN)
• A unified data catalog across domains
• SLAs for data freshness, quality, and accessibility

Data must become a first-class citizen in network operations and not an afterthought.

Closing thought: autonomy is about federated data product ownership and distributed intelligence

Autonomous networks are developed by re-architecting how data is managed, shared, and trusted.

A modern, cloud-native data lake ensures data is collected and stored on a scale. A robust data mesh ensures data is reliable, actionable, and owned where it matters. Together, they lay the groundwork for a network that can sense, decide, and act—without human intervention.

By adopting hybrid architectures, federated data ownership, and distributed intelligence, CSPs can unlock new levels of operational efficiency, customer experience, and innovation