Predicting article demand with Temporal Fusion Transformers
Picnic is the world's fastest growing online supermarket that makes grocery shopping simple, fun, and affordable for everyone. To ensure the freshest products and reduce waste, Picnic operates as a just-in-time supply chain. This must be balanced against high availability requirements for grocery items, as one unavailable product might lead to the loss of an entire basket. Accurate article demand forecasts are paramount. In this talk, we'll share how Picnic optimizes article demand forecasts with ML models. We'll dive deeper into why we transitioned from tree-based models to deploy the more state-of-the-art Temporal Fusion Transformer model and how it's used to balance waste & availability.
Sharon Gieske is the Tech Lead of the Data Science team at Picnic. Here, she leads the development of technologies that enable Machine Learning projects in a wide range of domains such as demand forecasting, personalization, finance and more. Together with her team of data scientist, she works full-stack and builds end-to-end ML solutions to take e-commerce at Picnic to the next level.