Bad data hurts model performance and invalidates accuracy metrics. Aquarium integrates with your labeling system to help you find and fix data quality issues.
Inspect your data and labels as you receive new batches of labeled data from your labeling system.
Quickly find errors in your labeling process and optimize your labeling spend.
Find errors quickly
Surface problematic examples so you don't have to manually inspect your entire dataset for errors.
Find examples with high disagreement between your model and your data - they’re most likely to have data errors.
“There are two ways Aquarium provides value to my company. First, we improved our model performance. Second, I spent less time and less clicks curating my dataset."
Tom Shapland, CEO of Tule Technologies
Solve problems together
Collaborate with ML engineers, data operations teammates, and other stakeholders to discuss and resolve data problems.
Share links to interesting subsets of your dataset with your team.
Assign issues to teammates, leave detailed comments, and track resolution of problems.
Speed up your iteration cycles
Set up a tight feedback loop with your labeling system with Aquarium's integrations.
Periodically pull in new data as it’s labeled with streaming dataset updates.
Easily export data to your labeling system via webhooks and out-of-the-box API integrations for relabeling.
