Predibase: A low-code deep learning platform built for scale

Predibase is a new kind of low-code machine learning platform from the creators of Ludwig and Horovod, combining large-scale deep learning with state-of-the-art model architectures across a range of modalities, including NLP, computer vision, tabular, and hybrid data. Unlike existing AutoML solutions, Predibase provides a modular "lego brick" experience that allows data scientists and engineers alike to iteratively build the best model for their data without needing to master a low-level framework like PyTorch or TensorFlow. In this talk, we'll explain how Predibase uses Ludwig and Ray under the hood to deliver an end-to-end, cloud-native solution encompassing data processing, model training, hyperparameter optimization, and model serving, all done at scale. We'll also explore how Predibase provides a unique serverless abstraction over Ray, allowing us to support multi-cloud and multi-tenant deployments capable of handling many concurrent users and requests.