Databricks Data Engineering Beneath Demand Intelligence

A global sportswear brand engaged Taller to build and operate the Databricks data-engineering substrate feeding its demand-intelligence and forecasting platform.

The client’s demand-intelligence platform (built with AWS Professional Services using DeepAR and SageMaker, Amazon’s forecasting and machine-learning tools) was only as good as the data feeding it: inventory, sales, and demand-planning data drawn from across the client’s wider data estate. Getting that layer right was the precondition for the whole machine-learning investment to pay off. The client’s architecture split the work into distinct layers: Databricks for data engineering and SageMaker for the model layer, with each depending on the one beneath it. The client engaged Taller to own that foundational layer.

Taller ran a Databricks data-engineering team for the client, working across the data estate in Python, SQL, and PySpark (a tool for processing very large datasets), building the ETL pipelines that extracted, cleaned, and reshaped data on an agile cadence. Based across Argentina and Colombia, this was a sustained data-engineering capability rather than a short-term project. Alongside the engineers, Taller also provided SAP finance and demand-planning business analysis (spanning the APO, IBP, and SAC modules), translating the business’s planning, forecasting, and budgeting needs into specifications the technical team could build against. The SageMaker models themselves belonged to the client’s internal machine-learning team working with AWS. Taller’s data engineering fed the foundation those models depended on, and its SAP business analysis shaped the demand-planning requirements flowing into them.