As published on wandb.ai, April 2024. Captured by usedby on Oct 3, 2026.
What happened
Canva's ML platform team uses Weights & Biases for experiment tracking and uses the W&B Registry as a central hub to move models from experimentation to deployment. The Registry separates production-ready models from experimental ones, and aliases show which model is deployed to production or used for A/B testing.
Summary written by usedby from the source page, in English. The figures are those of Weights & Biases and Canva, not ours.
Using the W&B Registry provided a clear separation between production and experimentation, while aliases made it much clearer for users to understand which model would be deployed to production or to carry out A/B testing.
Weave enables our engineers to score changes to product features using different LLMs under the hood. This allows them to make an informed decision about which one to use, taking into consideration not just quality but also cost.
The W&B Registry simplifies our lives in so many ways. It brings less noise to the user experience, as we are now only seeing models that are production-ready. It stores all the production-level information we need.
What the story claims, and what we checked
We compared the story with its live page on Oct 3, 2026.
- The passage quoted aboveCheckedCopied word for word from the page, near the name of Canva.
- The publication dateCheckedRead from the page’s own metadata, never guessed.
- Canva uses Weights & BiasesCheckedConfirmed line. Latest check across sources: Oct 3, 2026.
- The result itselfNot checkedWe quote it; we did not measure it.




