Microsoft and Google Back Apache Ossie for Semantic Model Interoperability
Microsoft and Google join Apache Ossie, an open specification project aiming to make semantic models interchangeable across enterprise data, analytics, and AI platforms.
Microsoft and Google have joined the Apache Ossie project, an initiative focused on building an open specification for exchanging semantic models across various analytics, data, and AI platforms. Originally starting as the Open Semantic Interchange before moving to the Apache Incubator in June, Ossie already enjoys backing from more than 60 organizations. Supporters include Databricks, Snowflake, Nvidia, Salesforce, Informatica, Mistral AI, Oracle, ThoughtSpot, Sigma, and Tableau.
The framework represents datasets, fields, relationships, metrics, and AI context using JSON and YAML. Rather than requiring distinct point-to-point converters for every combination of vendors, Ossie functions through a hub-and-spoke model. The format serves as a common semantic layer, while specific converters translate between the hub and individual vendor implementations.
Platform contributions and technical scope
Microsoft's participation involves building a two-way converter for Power BI semantic models and Ossie, letting business context defined in Power BI translate into the standard format and vice versa. Furthermore, Microsoft is advocating for the inclusion of its ontologies within the specification and pushing for DAX—the expression language utilized by Power BI for calculations and business metrics—to be recognized among Ossie's query languages. This addition would allow calculation logic to accompany the model during platform migrations.
Google is also in the process of joining Apache Ossie. While specific contribution plans have not yet been detailed, the Ossie specification already incorporates BigQuery and GoogleSQL as a supported dialect due to community contributions recognizing the platform's large enterprise footprint.
Industry experts note that backing from these two major technology firms could alleviate redundant semantic engineering when moving workloads between systems. According to Ashish Chaturvedi of HFS Research, treating metrics as versioned, reviewable code artifacts helps prevent metric drift compared to manually recreating business definitions on every platform. For developers building agentic workflows, maintaining a single definition across tools can lower the risk of artificial intelligence agents interpreting metrics differently, noted Stephanie Walter of HyperFrame Research. Developers can try top AI models cheaply through one API at https://apixoai.online.
What it means for developers
For engineering teams, interoperability standards like Ossie can reduce the time spent validating and translating semantic definitions when onboarding new tools or deploying AI applications. Michael Leone of Moor Strategy and Insights points out that increased portability grants enterprises greater ownership of their data models and more leverage when evaluating software buyers' options.
However, portability does not equal complete interoperability. Analysts caution that equivalent behavior is not guaranteed simply because a structure is portable. Destination platforms may interpret nulls, filters, joins, or time calculations differently. For instance, complex Power BI time-intelligence functions and DAX calculation groups do not always map cleanly to ANSI SQL, meaning sophisticated models might lose some functionality during translation.
Current limitations and vendor lock-in
Several gaps remain in the current development draft. Governance elements such as row-level security, access policies, and certification status are absent from the core specification, requiring enterprises to reconfigure them manually on each system. Additionally, the project is currently in the Apache Incubator stage at version 0.2, meaning the schema could change as it matures.
Analysts also suggest that Ossie is unlikely to completely erase vendor lock-in. While definitions may become portable, execution engines will likely maintain proprietary behaviors and vendor-specific extensions that interchange formats cannot fully capture, shifting the nature of platform dependency rather than eliminating it entirely.
Source: Microsoft, Google back Apache Ossie to make enterprise data and AI platforms more interoperable — InfoWorld AI. Written by the Apixo team from that report.
One key for Claude, GPT, GLM, DeepSeek and more. Pay per token with crypto.
Get your API key
