← All selected work
02

Data platform / modernization

From fragmented Azure resources to a unified Fabric platform

A full rebuild consolidated years of Azure sprawl into a governed Microsoft Fabric platform with lower cost, faster processing, and hourly operational visibility.

60%+

lower infrastructure cost

20 -> 3m

overnight batch runtime

Hourly

operational refresh

300+

users and leaders supported

Environment

Analytics for a manufacturing, reverse-logistics, and sales organization supported more than 300 team members across operations, finance, HR, sales, and executive leadership.

Constraint

Years of changing requirements and ownership had left Data Factory, Azure SQL, Blob Storage, Functions, Logic Apps, Power BI models, and reports fragmented and difficult to manage. Cross-department analysis was slow, refreshes lagged, and development happened directly in one production workspace.

01

S3 + SFTP + business systems

02

Fabric ingestion

03

Bronze + silver Lakehouses

04

dbt + SQL gold models

05

Semantic models

06

Reports + agents

01

Rebuild, do not just port

The migration moved through discovery, consolidation, and rebuild. Rather than reproducing legacy Data Factory jobs, Paul rewrote processing logic as scalable PySpark notebooks with parallel execution and a medallion architecture.

Native S3 shortcuts, SFTP labor feeds from Paycom, Business Central integrations, and internal sources converged in Fabric Lakehouses. dbt and SQL produced governed gold models for shared semantic and reporting layers.

02

Establish an operating model

The rebuilt platform introduced Git-backed source control, monitoring, separate development and production environments, and controlled deployment pipelines in place of a single shared workspace.

Key Vault protected secrets, service principals authenticated resource access, workspace roles governed Fabric, and Power BI row-level security separated financial information from broader operating audiences.

03

Move from daily to operational

Overnight batch processing fell from roughly 20 minutes to 3 minutes. A parallel hourly path refreshed the same S3 operational data throughout the day, giving leaders current visibility without sacrificing the governed batch layer.

The completed migration retired obsolete Azure resources, reduced monthly infrastructure cost by more than 60%, and maintained near-continuous availability across reporting and infrastructure.

04

Add an intelligent access layer

Production Microsoft Foundry agents connected to governed Fabric data through Data Agents, Fabric MCP, and custom tools. The platform could answer plain-language questions while retaining the semantic, identity, and access controls established underneath it.

Paul's ownership

Paul identified the backend as the reporting constraint and independently led discovery, architecture, stakeholder alignment, implementation, migration, modeling, Power BI delivery, governance, deployment, training, and ongoing operation over a roughly one-year program.

Capabilities

  • + Platform architecture
  • + Data engineering
  • + Migration
  • + Governance
  • + BI delivery
  • + FinOps

Selected stack

  • + Microsoft Fabric
  • + PySpark
  • + dbt
  • + SQL
  • + Power BI
  • + Azure Key Vault
  • + Git
  • + S3

Next system / 03

Turning analyst requests into governed, self-service answers

Start with the problem

Have a data or AI system worth untangling?