Noether

The Noether Framework is an open source deep learning framework for engineering AI, built by Emmi AI. It supplies the pieces needed to build, train and apply industrial models on transformer building blocks, so engineering teams do not have to rebuild components or assemble an in house deep learning team. The framework covers three areas. Data management handles high dimensional physics datasets from storage such as HuggingFace or AWS S3, with building blocks for preprocessing. Training is fully configurable through YAML files, supports distributed GPU clusters, mixed precision and dynamic scheduling, and includes ready to use architectures for data such as CFD meshes and volumetric fields. Inference is optimized for trained models, and Emmi AI also offers a growing library of pre trained physics models for simulation tasks. It installs with pip or from the GitHub repository, and documentation and examples are linked from the page. The target users are simulation and engineering groups in areas such as aviation, energy, semiconductors and automotive. The framework itself is free to use. For commercial deployments, Emmi AI offers custom model training on proprietary datasets, enterprise support and SLAs, integration with existing simulation pipelines, performance optimization and training workshops, and these are arranged by contacting the team. No prices are published. The page notes that Emmi AI has joined Mistral AI.