NitroML is a framework for benchmarking Automated Machine Learning (AutoML) pipeline model-quality.

Our mission is to provide model-quality benchmarking tools to accelerate AutoML research and development.

More generally, NitroML enables AutoML research teams to iterate more quickly on their highly-customized, heterogeneous, and multi-stage pipelines. It offers machine learning model-quality benchmarking best practices out-of-the-box, curates raw datasets, and leverages [TFX:OSS] to scale with Cloud-resources. Its benchmark database and analysis tools ensure that AutoML teams can be data-driven as they modify their systems.

This is not an officially supported Google product.

For more information, see go/nitroml.