Databricks Data Engineering Beneath Demand Intelligence

The client is the second-largest global sportswear manufacturer and the company behind the "their digital transformation strategy" strategy that committed over a billion euros to digital transformation through 2025. The 2025 results land at tens of billions in net sales, billions in direct-to-consumer revenue, and multi-billion operating profit with strong e-commerce growth. Their loyalty program with over 300M members operates across roughly fifty countries; e-commerce runs in sixty-five. Taller's engagement at this client has run continuously since September 2021, sourced through a staffing partner. The dominant workstream is the North America intelligent process automation program on UiPath — captured in Volume II under the RPA Automation Anywhere case with $2M+ annual cost savings. Adjacent workstreams cover Power BI analytics for the client's tech delivery hub (also captured in Volume II), an eCommerce/B2B/Retail platform-engineering pod under the client's tech delivery hub, Databricks data engineering, SAP integration and functional business-analysis capability, EDI engineering for B2B trading, and the Latin America tech infrastructure team service-management team in Brazil and Mexico. The three validated use cases below land in three of these adjacent surfaces.

The demand-intelligence platform the global sportswear manufacturer runs with AWS Professional Services on the DeepAR/SageMaker stack requires a data substrate that the ML model can consume - clean, timely, well-modeled inventory, sales, and demand-planning data across the broader global sportswear manufacturer data estate. The model layer is only as accurate as the substrate it operates on, and getting the substrate right is the engineering precondition for the broader ML investment to compound. The global sportswear manufacturer's data architecture choice (Databricks for the data engineering substrate, SageMaker for the model layer) creates a layered architecture where each layer's discipline depends on the previous one.

Taller operates a Databricks data engineering team at this client (Python, SQL, PySpark, ETL discipline on Agile cadence), working on data pipelines across the broader data estate. The talent base sits across Argentina and Colombia and represents a sustained data-engineering capability buildout rather than a one-off project. Adjacent to the data engineers, Taller covers SAP Finance/Demand-Planning functional business analysis spanning APO, IBP, and SAC functional knowledge — the layer that translates business requirements for demand planning, financial planning, forecasting, and budgeting into specifications the technical team operates against. Taller does not write or operate the SageMaker DeepAR models themselves - those are owned by the global sportswear manufacturer's internal ML team in partnership with AWS Professional Services - but Taller's data engineering work feeds the substrate the models consume, and the SAP functional business analysis work shapes the demand-planning requirements that flow into the analytical layer.

Continuous Databricks data engineering operation across the global sportswear manufacturer data estate. Client feedback through 2024 and 2025 references the team's sustained effectiveness; the data engineering substrate operates against the analytical workstreams that demand intelligence depends on.

The dossier's strategic frame on UC2 is demand intelligence as the lever that converts probabilistic forecasting into working-capital discipline (the forty-percentage-point reduction in over-prediction error vs. baseline). Taller's contribution is at the data substrate and functional-BA layer that the analytical layer rests on. The data engineering work compounds with the ML investment by ensuring the model has clean, timely, well-modeled data to operate against; the SAP functional BA work compounds by ensuring the demand-planning requirements are translated cleanly across business and technical boundaries.