Ludwig version 0.4, an open-source, low-code declarative deep learning framework by Uber, introduces MLOps best practices to enhance scalability in data processing, training, and hyperparameter search. New integrations include Ray for distributed training, Ray Tune for hyperparameter search, and MLflow for experiment tracking and model serving. The release adds TabNet as a combiner for tabular data, and preconfigured datasets from Kaggle are now available for easier experimentation. Ludwig's declarative approach simplifies machine learning pipelines by allowing users to configure them via YAML, enabling flexibility and ease of use. The integration with Ray allows users to scale machine learning tasks from local environments to cloud-based multi-node clusters without code changes. Future developments include features like end-to-end AutoML and a new enterprise platform, Predibase, to further simplify machine learning processes.