Instacart, an Internet-based grocery delivery service, shares how they are using deep learning to help their tens of thousands personal shoppers be more efficient.
“By observing how our shoppers have picked millions of customer orders through our app, we have built models that predict the sequences our fastest shoppers will follow,” mentions Instacart VP of Data Science Jeremy Stanly in a blog that details a variety of the different deep learning-based approaches they explored. “Then, when a shopper is given a new order to pick, we use this predicted fastest sequence to sort the items for them.”
Using Tesla K80 GPUs on the Amazon cloud, Keras and TensorFlow deep learning framework to train their models, the team’s final architecture produced significant performance and efficiency gains which has reduced their shopping times by minutes per trip.
The team’s final method is a ‘scoring generator’-based architecture that helps suggest the next candidate depending on the prior product picked.
Read more >
Related resources
- GTC session: Connect with the Experts: Deep Learning, Machine Learning, and Data Science (Spring 2023)
- GTC session: Building Session-Based Recommendation Models with NVIDIA Merlin* (Spring 2023)
- SDK: Merlin
- SDK: Merlin HugeCTR
- Webinar: How AI Is Reinventing the Retail Industry
- Webinar: NVIDIA NGC Jupyter Notebook Day: Recommender System
About the Authors
