Case Study: Shipt scales ML operations with Comet

How Shipt Leverages Comet

Shipt, the grocery delivery service serving millions of U.S. households, needed a way to scale machine learning operations across a multi-sided retail marketplace while keeping its ML platform flexible for data scientists and production teams. The company’s challenge was managing experimentation, artifacts, and model promotion across development and deployment without forcing a monolithic, all-built-in-house platform.

With Comet, Shipt logs training runs from Airflow, JupyterHub, and local environments using the Comet Python SDK, then promotes the most promising models into Comet’s Model Registry for deployment. This hybrid approach lets Shipt focus on core platform work while relying on Comet for experiment tracking and registry management, helping streamline model promotion, improve consistency from training to production, and support a more robust ML deployment process.

Shipt

Adam Hendel

Principal Machine Learning Engineer