Croissant: a metadata format for ML-ready datasets

Machine learning (ML) practitioners looking to reuse existing datasets to train an ML model often spend a lot of time understanding the data, making sense of its organization, or figuring out what subset to use as features.