What we actually do.

The foundation everything else sits on. Modelled deliberately, indexed on evidence, and documented well enough that the next person can pick it up.

• Dimensional and Data Vault modelling, with the trade-offs explained before anything is committed to
• Query and index tuning against real execution plans, not guesswork
• Partitioning, compression and storage strategy for tables that outgrew their design
• Historical loads and slowly changing dimensions that survive a restatement

One semantic model the business agrees on, refreshing inside its window, with security that holds when someone shares a link.

• Fabric lakehouse and warehouse workloads, with capacity sized against real usage
• Semantic models built for performance: correct grain, tuned relationships, sane DAX
• Row level and object level security that survives an audit
• Deployment pipelines so development, test and production stop drifting apart

The work most teams put off. Done incrementally, with the old system running alongside until the new one has proven itself.

• SSRS to Power BI, including paginated reports that genuinely need to stay paginated
• On premises SQL Server to Azure SQL, Managed Instance or Fabric, chosen on cost and constraint
• Serverless architectures where the workload is spiky, provisioned where it is not
• Reconciliation harnesses that prove old and new agree before anything is switched off

Lakehouse platforms for teams whose data outgrew a single server, built with the same discipline as a warehouse.

• Medallion architecture with contracts between layers, not just folder names
• PySpark and SQL transformation layers under version control and test
• Snowflake warehouse sizing, clustering and cost controls that hold up month to month
• Streaming and micro batch ingestion where latency actually matters to someone

Pipelines that fail loudly, retry sensibly, and can be rerun for a single day without a support ticket.

• Metadata driven pipelines that add a new source through configuration rather than copy and paste
• SSIS packages brought into source control, or retired onto Data Factory and Fabric
• Idempotent loads with proper watermarking, so a rerun is safe
• Architecture reviews with a written recommendation you can take to a budget holder

Models and automation that reach production, plus the delivery pipeline that keeps the rest of the platform honest.

• Forecasting, classification and anomaly detection trained on your warehouse, not a sample extract
• Document extraction and retrieval over internal content, with the retrieval layer built properly
• Model tracking, versioning and monitoring so you know when accuracy drifts
• Azure DevOps pipelines for database projects, Power BI artefacts and infrastructure as code