An open platform for end-to-end foundation-model training, evaluation and deployment across text and vision models.