Gravitino is a high-performance, geo-distributed, and federated metadata lake (github.com)

Apache Gravitino is a new Apache metadata lake designed to provide high-performance, geo-distributed, and federated metadata management for data and AI assets. It exposes a single model and API to unify metadata from heterogeneous sources (Hive, MySQL, HDFS, S3) and supports federated discovery across data lakes and warehouses. Gravitino is built for global architectures: multi-region synchronization and cross-cloud sharing let teams maintain a consistent metadata view for hybrid and multi-cloud deployments. For governance-sensitive environments it offers unified access control, auditing, and discovery, and it’s explicitly adding AI asset features like model and feature tracking and lineage support. Technically, Gravitino integrates directly with underlying systems via connectors so metadata changes are reflected immediately, and it provides a native Apache Iceberg REST catalog plus a Trino connector for plug-and-play query-engine access (Spark/Trino) without changing SQL dialects. It targets high performance and federation, is built with Gradle (Windows not yet supported), and is available under the Apache 2.0 license. For engineers and ML platform teams, Gravitino promises to simplify cross-region metadata consistency, enable unified governance of models and datasets, and reduce friction when connecting multiple query engines and storage backends.