Nuclear Waste Services: A Governed Azure Data Platform Built on Medallion Architecture

Overview

One trusted data platform for a nuclear-sector organisation

Nuclear Waste Services (NWS) is part of the Nuclear Decommissioning Authority group and is responsible for the safe disposal of the UK's radioactive waste. Its programmes depend on accurate information about schedules, resources and costs.

That information lived in separate business systems. Project planning data sat in an enterprise project portfolio management system, and financial data sat in an ERP system. Bringing it together for reporting meant extracting data system by system, with limited visibility of how figures had been derived.

Konsolute delivered an Azure data platform based on a medallion architecture. Data is ingested through each system's APIs, landed in its raw form and progressively cleansed, conformed and curated. Pipelines are orchestrated with Azure Data Factory and Azure Synapse Analytics, and changes are managed and released through Azure DevOps.

The challenge

Valuable data, held in disconnected systems

NWS needed a single, governed source of programme and financial data. Reporting had to draw on several systems at once, and each source had its own data model, interface and refresh pattern.

In a nuclear-sector organisation, it is not enough for reports to look right. Teams need to trace a figure back to its source, understand what happened to it on the way, and rely on data being handled securely. Any new platform had to make lineage, quality and access part of its design.

The scale of the data was also unknown at the outset. Data volumes, record counts and rates of change had to be understood before storage, pipeline schedules and load patterns could be designed sensibly.

NWS needed to:

Our approach

Measure the data, then build it layer by layer

We began with the data itself: what each system held, how much of it there was and how quickly it changed. That evidence shaped every layer of the platform.

Discovery and volumetric analysis. Konsolute profiled each source system, measuring entity and record counts, payload sizes and rates of change. The volumetrics informed storage design, pipeline schedules and the choice between full and incremental loads for each entity. They also gave NWS a basis for planning capacity and cost as the platform grows.

API-led ingestion. Data is extracted through each system's APIs, including REST and OData interfaces, using secured, token-based authentication. Pipelines page through large result sets, handle failures and retries, and record each run so that loads can be monitored and reconciled.

Bronze layer. Raw data lands in Azure Data Lake Storage Gen2 exactly as received, in JSON and Parquet formats. Keeping an unaltered copy preserves an audit trail and lets data be reprocessed when business rules change, without returning to the source system.

Silver layer. Data is cleansed, typed, deduplicated and conformed to consistent definitions. Relationships between entities, such as projects, activities and resources, were modelled so that data from different systems can be joined reliably. The ERP system was added as a parallel source alongside the project planning system, following the same patterns.

Gold layer. Curated, business-ready datasets are shaped around reporting needs. This is the layer analysts and reports consume, with agreed definitions applied once rather than rebuilt in every report.

Synapse and Data Factory infrastructure. Azure Data Factory and Synapse pipelines orchestrate extraction and transformation across the layers. The platform's Azure infrastructure was configured with access controls appropriate to NWS's security requirements.

DevOps and controlled release. Pipelines, transformation logic and configuration are held in source control in Azure DevOps. Changes move through review and release pipelines rather than being edited directly in production, giving NWS a repeatable and auditable route for platform change.

Throughout, Konsolute produced formal technical design documentation covering the architecture, data flows, entity models and operating procedures, so that NWS teams can understand, support and extend the platform.

Result

A foundation for reporting the organisation can trust

NWS now has a single Azure data platform that brings programme and financial data together. Data flows from source APIs through bronze, silver and gold layers, so every curated figure can be traced back to the raw data it came from.

The volumetric work means the platform was designed around real data volumes rather than assumptions, with load patterns matched to how each entity changes. Retaining raw data in the bronze layer allows NWS to reprocess history when definitions change.

Because changes are managed through Azure DevOps, the platform can evolve in a controlled way. New sources follow established ingestion and modelling patterns, as the addition of the ERP system alongside the project planning system demonstrated.

The platform gives NWS governed foundations for reporting today, and for analytics and AI use cases in future.