Data Warehouse Modernisation Lead

Job description

BCB is seeking a contractor to own the design, build and operation of the bank’s new data warehouse on Microsoft Fabric: the domain model, the ingestion pipelines from our core banking platform, the semantic layer, and the security and governance that make the data trustworthy in a regulated environment. This is a hands-on lead contract role that will work as an embedded member of the bank’s team, delivering one business area from source system through governed, tested data to a live report before moving to the next. Key Accountabilities and Responsibilities Platform and warehouse build Build and run the Microsoft Fabric warehouse: workspaces, capacity management, OneLake structure, and bronze/silver/gold layers. Design and own the business domain model (customers, accounts, balances, transactions) — modelling the domain properly rather than cloning the legacy reporting estate. Untangle and document the legacy reporting paths: our T24 core banking platform feeds both the accounting software and the existing reporting tool through different, undocumented routes that carry the same underlying data. Ingestion Build Fabric-native ingestion for phase one: Data Factory pipelines, Mirroring, Dataflows Gen2 and OneLake Shortcuts. Design bespoke extraction from the on-premise core banking source — this is a legacy platform, not SaaS, so hand-built extraction is expected regardless of tooling. Own the decision on when to introduce a managed ingestion tool such as Fivetran, or reverse-ETL, against agreed triggers for revisiting. Modelling and the semantic layer Own the call between dbt and native Fabric modelling (stored procedures, T-SQL views, notebooks) for the gold layer, and mitigate whichever gaps that choice leaves — documentation and tests staying current, and versioning discipline via Git integration and pull-request review. Build and own the single authoritative Fabric semantic model — the one place metric definitions live for Power BI, Excel, XMLA-based tools and AI access alike — and keep direct SQL/table access locked down so it cannot be bypassed. Security, entitlement and governance Design and implement least-privilege access with row-level and column-level security enforced at the data layer, not the BI layer. Own and re-run the two-user entitlement test as a mandatory gate before any AI or BI tool goes live against customer data, and after any change to connectors, roles or security configuration. Produce the data dictionary and lineage the bank’s audit work needs, reconcile every Finance figure back to source, and build quality tests as code from the first delivery slice. AI enablement Own the Fabric Data Agent (or equivalent) setup that lets approved AI tools query the semantic model via delegated user identity, never a shared service account. Design testing that reflects the non-deterministic nature of AI query composition (repeated runs, varied phrasing) while security enforcement remains deterministic. Use AI as an accelerator to reverse-engineer and document legacy reporting logic — modelling the business domain from it, never cloning flawed reports. Regulatory awareness Work with the DPO and Compliance on the DPIA before any personal data enters the pipeline; keep AI-facing data to schema, report logic and masked samples by default. Design with the EU AI Act’s high-risk obligations in mind for the future AI-assisted transaction-monitoring and payment-screening workstream, without owning compliance sign-off personally. Experience, Skills & Attributes Deep, current Microsoft Fabric expertise: OneLake, Data Factory, Dataflows Gen2, semantic models, Direct Lake, and row-level/column-level security. Comparable depth in Azure Synapse plus Power BI Premium will be considered where Fabric exposure is more recent. Breadth across the modern data stack. We are not looking for someone who knows every tool, but your track record should show you have worked hands-on with several: at least one other warehouse technology (for example Databricks, Snowflake, BigQuery or Redshift) and several tools from the wider stack, such as dbt for modelling, a managed ingestion tool like Fivetran or Airbyte, or orchestration tooling. Strong SQL / SQL Server background, comfortable owning both pipeline and modelling work rather than only one side. Demonstrable ability to reverse-engineer an undocumented legacy reporting system; direct experience of core banking platforms (ideally T24) is a strong advantage. Solid understanding of data governance in a regulated environment: lineage, data dictionaries, quality testing as code, and reconciliation to source. Would be an asset to have… Working knowledge of dbt. Exposure to AI/LLM-integrated data architectures (semantic layers, agent-based query tools). Experience in banking or another regulated financial services environment.