Microsoft and Google will contribute to Apache Ossie, an open-source project aimed at making semantic models interchangeable across different data, analytics, and AI platforms. This adds two major participants to an initiative already supported by dozens of large tech companies.
Ossie aims to address a problem that grows as organizations combine more data sources, analytics tools, and AI applications. A semantic model defines what data means within an organization and how it relates to other data, as InfoWorld explains. Think, for example, of the definition of revenue, customer, or gross margin. Such definitions are currently often tightly tied to the platform on which they were created.
This makes it difficult to use the same business logic across different environments. Ossie aims to provide a common middle layer to address this.
A single format across different platforms
The initiative began as Open Semantic Interchange (OSI) and was accepted into the Apache Incubator in June. Since then, it has been known as Apache Ossie. Supporters of the project include Databricks, Informatica, Mistral AI, Nvidia, Oracle, Salesforce, and Snowflake.
Ossie describes semantic models in JSON and YAML. These models can define datasets, fields, relationships, and metrics, as well as the context that AI systems need to interpret data correctly.
The approach does not revolve around a single universal data platform. Instead, Ossie acts as an intermediate format. Converters translate the semantic models from individual products to Ossie and back again. This eliminates the need to build a separate integration for every combination of two platforms.
This can be particularly relevant for organizations that combine Snowflake or Databricks with analytics tools such as Tableau, ThoughtSpot, and Sigma.
Microsoft is building a Power BI converter
Microsoft is working on a bidirectional converter for Power BI. This should enable a semantic model from Power BI to be translated into Ossie and then used in another compatible environment. Conversely, Ossie models can be converted back to Power BI.
The company also wants to see elements of its own technology incorporated into the standard. Among other things, Microsoft advocates for more extensive support for ontologies and for recognition of DAX. This language is used within Power BI to define calculations and business metrics.
This is precisely where the challenge for Ossie lies. Vendors can support a common standard while simultaneously using features specific to their own platform. Full portability is therefore not a given.
AI heightens the importance of semantics
The rise of generative AI adds further weight to this issue. AI agents and other systems that independently consult business data need more than just access to tables and columns. They also need to understand the meaning of terms and the relationships and business rules associated with them.
A shared semantic layer can prevent the need to rebuild that context from scratch for every AI or analytics environment. This will ultimately allow Ossie not only to simplify the migration of analytics models but also to play a role in making reliable business context available for AI.
Whether this actually succeeds depends on the further development of the specification and on the extent to which vendors can translate their own semantic features into the common model without loss of information. In any case, the broad support from the industry makes Ossie a project worth keeping an eye on.