More
Call to Collaboration
To Know
-
PROJECT STORIES
From Pilot to Platform
Reshaping our client's network operations
Call to Collaboration
—
Sep 2026
Our client is pursuing a multi-year transformation of its network operations in Romania, moving fault management, monitoring and orchestration onto Celfocus's Autonomous Network Operations program. Celfocus proposed proving its value first, quickly, and then scaling with confidence.
The pilot, which kicked off in May 2026, aims to deliver cross-domain fault management across RAN and microwave transmission, while establishing a platform baseline designed to support the wider program. This marks the start of a multi-year journey running through 2027 and beyond, ultimately bringing the client's full Romanian network estate onto the Autonomous Network Operations program.
The Autonomous Network Operations program runs on Google Cloud Platform and is structured in five layers.
Observe handles network mediation, ingesting 1) real-time alarms from the client's Fault Management Umbrella (FMU) through a purpose-built FMU adaptor and 2) network topology data extracted from the client's inventory and OSS systems.
Data persists raw, standardised and curated data in the appropriate storage services, harmonizing, grouping, and exposing that information through a common layer - including a graph database - to support topology traversal, analytical processing and network digital twin use cases. Decide is the AI/ML-driven decision engine that performs cross-domain root-cause analysis, correlating RAN and microwave transmission alarms, filtering noise and automatically classifying alarms, backed by an MLOps pipeline for inference and ongoing model management.
Act converts AI insights into orchestrated workflows, applying business rules and domain policies to execute predefined workflows associated with the outcomes produced by Decide layer.
A Unified Operations Portal sit across all four, giving operators a single point of monitoring, control and analytics.
The pilot itself is being delivered in three structured phases:
The pilot closes with a three-week, client-led validation phase in a pre-production environment running real production data.
The pilot delivers measurable operational impact from day one: cross-domain correlation across RAN and microwave transmission for unified fault management with root-cause analysis, ML-driven noise filtering and alarm classification to cut mean-time-to-detect and mean-time-to-act and a single operations portal with analytics dashboards for end-to-end visibility. Progress is tracked against a defined set of business outcomes.
Just as importantly, nothing built for the pilot is single-use. The GCP-ready platform baseline - with the Observe, Data and Decide layers, Portal and orchestration/workflow management already pre-integrated is designed to carry directly into the full Romania rollout, with initial configurations for data, decision logic and topology ready to be reused rather than rebuilt.
By proving the platform on a contained, well-defined scope first, Celfocus reduces delivery risk on a multi-year, multi-domain program while building reusable assets that extend well beyond Romania. The same GCP-ready baseline is designed to accelerate deployment for the client's other GCP-based operators, turning a single-country pilot into a repeatable blueprint. The engagement also reinforces Celfocus's Autonomous Networks proposition, combining multi-domain and multi-layer operations, digital twin thinking and agentic AI.
The pilot is delivered by a dedicated, cross-functional Celfocus team organized across two levels. A steering group comprising Celfocus account and delivery leadership manages the overall program across multiple affiliates at group level. For each operator, an execution team brings together the technical core, architecture, project management, data, AI and infrastructure leads, and a process analyst, with each role paired with a client counterpart from the respective affiliate.
With the pilot well underway through the second half of 2026, it will next move into a three-week, client-led validation phase. Meanwhile, discussions are progressing on continuing the project to deliver the full scope and onboarding the next operators. The same reusable components position the Autonomous Network Operations program for rollout across the client's other operators, ahead of an eventual handover of operations and support to the client's own teams.
This article is brought to you by the Benelux & UK Account Cluster team.
Our client is pursuing a multi-year transformation of its network operations in Romania, moving fault management, monitoring and orchestration onto Celfocus's Autonomous Network Operations program. Celfocus proposed proving its value first, quickly, and then scaling with confidence.
The pilot, which kicked off in May 2026, aims to deliver cross-domain fault management across RAN and microwave transmission, while establishing a platform baseline designed to support the wider program. This marks the start of a multi-year journey running through 2027 and beyond, ultimately bringing the client's full Romanian network estate onto the Autonomous Network Operations program.
The pilot is delivered by a dedicated, cross-functional Celfocus team organized across two levels. A steering group comprising Celfocus account and delivery leadership manages the overall program across multiple affiliates at group level. For each operator, an execution team brings together the technical core, architecture, project management, data, AI and infrastructure leads, and a process analyst, with each role paired with a client counterpart from the respective affiliate.
Our client is pursuing a multi-year transformation of its network operations in Romania, moving fault management, monitoring and orchestration onto Celfocus's Autonomous Network Operations program. Celfocus proposed proving its value first, quickly, and then scaling with confidence.
The pilot, which kicked off in May 2026, aims to deliver cross-domain fault management across RAN and microwave transmission, while establishing a platform baseline designed to support the wider program. This marks the start of a multi-year journey running through 2027 and beyond, ultimately bringing the client's full Romanian network estate onto the Autonomous Network Operations program.
The Autonomous Network Operations program runs on Google Cloud Platform and is structured in five layers.
Observe handles network mediation, ingesting 1) real-time alarms from the client's Fault Management Umbrella (FMU) through a purpose-built FMU adaptor and 2) network topology data extracted from the client's inventory and OSS systems.
Data persists raw, standardised and curated data in the appropriate storage services, harmonizing, grouping, and exposing that information through a common layer - including a graph database - to support topology traversal, analytical processing and network digital twin use cases. Decide is the AI/ML-driven decision engine that performs cross-domain root-cause analysis, correlating RAN and microwave transmission alarms, filtering noise and automatically classifying alarms, backed by an MLOps pipeline for inference and ongoing model management.
Act converts AI insights into orchestrated workflows, applying business rules and domain policies to execute predefined workflows associated with the outcomes produced by Decide layer.
A Unified Operations Portal sit across all four, giving operators a single point of monitoring, control and analytics.
The pilot itself is being delivered in three structured phases:
The pilot closes with a three-week, client-led validation phase in a pre-production environment running real production data.
The pilot delivers measurable operational impact from day one: cross-domain correlation across RAN and microwave transmission for unified fault management with root-cause analysis, ML-driven noise filtering and alarm classification to cut mean-time-to-detect and mean-time-to-act and a single operations portal with analytics dashboards for end-to-end visibility. Progress is tracked against a defined set of business outcomes.
Just as importantly, nothing built for the pilot is single-use. The GCP-ready platform baseline - with the Observe, Data and Decide layers, Portal and orchestration/workflow management already pre-integrated is designed to carry directly into the full Romania rollout, with initial configurations for data, decision logic and topology ready to be reused rather than rebuilt.
By proving the platform on a contained, well-defined scope first, Celfocus reduces delivery risk on a multi-year, multi-domain program while building reusable assets that extend well beyond Romania. The same GCP-ready baseline is designed to accelerate deployment for the client's other GCP-based operators, turning a single-country pilot into a repeatable blueprint. The engagement also reinforces Celfocus's Autonomous Networks proposition, combining multi-domain and multi-layer operations, digital twin thinking and agentic AI.
The pilot is delivered by a dedicated, cross-functional Celfocus team organized across two levels. A steering group comprising Celfocus account and delivery leadership manages the overall program across multiple affiliates at group level. For each operator, an execution team brings together the technical core, architecture, project management, data, AI and infrastructure leads, and a process analyst, with each role paired with a client counterpart from the respective affiliate.
With the pilot well underway through the second half of 2026, it will next move into a three-week, client-led validation phase. Meanwhile, discussions are progressing on continuing the project to deliver the full scope and onboarding the next operators. The same reusable components position the Autonomous Network Operations program for rollout across the client's other operators, ahead of an eventual handover of operations and support to the client's own teams.
This article is brought to you by the Benelux & UK Account Cluster team.
.png)
