Leverage unsupervised learning to improve supervised learning. Surface clusters of labeled data that your model struggles on and search within your unlabeled data to collect examples that improve performance on those issues.
Automated workflows speed up your model improvement by focusing human attention on the most critical scenarios. Don’t waste time manually trawling through data.
Embeddings made easy
Fully managed for you: Embeddings can be tricky to get started with on your own. Aquarium handles the infrastructure for generating, processing, and querying embeddings so you can quickly find improvements to your model performance.
Scale effortlessly to hundreds of millions of examples with Aquarium’s cloud based infrastructure. Built for supporting production workflows for ML teams.
Fine-tuned for your domain
Embedding workflows aren’t equal. Aquarium fine-tunes embeddings that are specific for your unique data domain, capturing important types of similarity better than embeddings extracted from pretrained models.
Few shot learning technology makes it easy to bootstrap new classes or source specific data with only a handful of examples.