The MLOps Platform has a built-in multi-model data layer (a.k.a. "data fabric", "data store", or "database") for storing and analyzing various types of data structures — such as NoSQL ("key-value") tables, data streams, binary objects, and files. Data is stored as objects within containers. The data objects can represent any supported data type (such as files, table items, or stream records), and objects can be grouped into type-specific collections (such as stream shards, NoSQL tables, or file-system directories).
The platform exposes and supports multiple industry-standard and industry-compatible programming interfaces that allow you to perform high-level data manipulation for the supported data formats. You can optionally switch between different APIs for accessing the same data; for example, you can ingest data through one interface and consume it through another interface, depending on you preferences and needs. You can often also access the same data in different formats. The platform's unique unified data model eliminates the need for multiple data stores, constant synchronization, complex pipelines, and painful extract-transform-load (ETL) processes. For more information, see The Data-Layer APIs.