Data engineering for ML-driven demand forecasting at Zalando

The challenge

Zalando's demand prediction and price optimization ran on heuristics that needed constant manual tuning. The data science teams wanted to move to modern machine learning, but they lacked a scalable, reliable data foundation to build on.

What I did

I owned the data engineering side: designing, building, and testing the data infrastructure that several data science teams depended on. That meant dependable pipelines, efficient processing of large-scale transactional and behavioral data across 500K+ products, and a foundation that ML models could be trained and deployed on without friction.

Outcomes

• Gave ML teams a high-performance foundation for model development and deployment.
• Helped replace manual heuristics with machine-learning-driven pricing.
• Supported Zalando's shift toward dynamic, AI-driven pricing at scale.
• Reduced repetitive manual work for the wholesale teams.