Case Study: Canva streamlines AI model management and deployment with Weights & Biases

Designing Streamlined Model Management Workflows at Canva Using Weights & Biases

Canva, the enterprise design and publishing platform with more than 150 million monthly active users, needed a better way to manage its growing machine learning workflows. Its ML Platform team wanted a cleaner separation between experimental runs and production-ready models, since the existing deployment process was noisy and relied on a complex set of tags. Canva used Weights & Biases, including W&B Model Registry, to help streamline model management and deployment.

With Weights & Biases Model Registry, Canva created a centralized hub for experiment tracking and production model management, making it easier to identify production-ready models and understand which versions should be deployed or used for A/B testing. The result was less noise in the user experience, clearer model governance, and access to the production-level information the team needed. Canva also reported that W&B improved usability across the team, with helpful system metrics and easier admin-side access and security management.

Canva

Thibault Main de Boissiere

ML Platform Team Lead