AtScale adds semantic layer support for AI, GenAI models
AtScale on Wednesday unveiled the latest version of its semantic layer platform, which features a new architecture that aims to enable decision-making agility and collaboration and includes support for both traditional AI as well as large language models.
In addition, AtScale unveiled a free version of its platform designed to introduce new users to the vendor’s sematic layer capabilities.
The vendor revealed the Developer Community Edition and overall platform update during the Semantic Layer Summit, AtScale’s annual user conference in Boston. Both are now in public preview.
Based in Boston, AtScale is a semantic layer specialist whose platform aims to help foster self-service analytics, data sharing and collaborative decision-making.
Semantic layers are tools that enable data administrators to define key metrics and standardize terms across their entire organization. The result is data consistency no matter what department is working with data and avoidance of data duplication.
With semantic layers, self-service users can find and share data without needing to know how to code or query and join data tables and sources. Essentially, they are a conduit for business users, according to Stephen Catanzano, an analyst at TechTarget’s Enterprise Strategy Group.
“The semantic layer serves as a bridge between raw data sources and end-user applications, providing a simplified, consistent, and efficient way to access and analyze data,” he said.
Meanwhile, they are especially beneficial now as the volume and complexity of data continue to increase, according to Kevin Petrie, an analyst at BARC U.S.
“This layer matters more now than ever because enterprises need to integrate and analyze increasingly diverse datasets across increasingly distributed, heterogeneous environments,” he said. “BI and data science teams need to reconcile different formats and schemas so they can analyze the data in a unified way.”
In addition to AtScale, vendors offering semantic layer capabilities include DBT Labs and Google-owned Looker, among others.
New capabilities
AtScale’s platform update centers on three pillars, according to the vendor.
One is flexibility with a Universal Semantic Hub to support all AI and BI platforms and enable organizations to work with data and AI products developed with any vendor’s tools. Another is collaboration with improved discoverability capabilities and both code-first and no-code modeling tools that encourage sharing between different personas within organizations. The last is community by enabling all users to build and share semantic models, including sharing between partner organizations.
To support flexibility, collaboration and community, AtScale’s update includes the following features within its semantic layer platform:
- A metadata hub that helps users discover the data needed to enrich both traditional and generative AI models with proprietary data so the models understand a given business and can provide domain-specific insights that inform decisions.
- Container Deployment, a feature that adds deployment options such as Kubernetes and Docker that are tailored for cloud-based customers and add scalability and automation capabilities.
- An integration with DBT Labs that lets DBT semantic models work with AtScale’s support for analytics platforms including Tableau, Microsoft Power BI and Microsoft Excel.
- A new Integrated Development Environment (IDE) that enables both code-first and no-code semantic data modeling and includes support for a continuous integration/continuous delivery integration with Git.
- Improved support for enterprise integrations and authentications through open source platforms Keycloak, OpenTelemetry, OpenAPI and KeyGen.
The emphasis on sharing, in particular, is noteworthy, according to Petrie.
“AtScale correctly recognizes the need to share data views across the enterprise and between enterprises,” he said. “By providing data teams with modular, portable, and reusable semantic objects, they help stakeholders across partner organizations standardize how they share data. This leads to more efficient collaborative analytics.”
In addition, the new IDE is a significant addition, Petrie continued.
“The new UX for semantic modeling should help drive adoption of semantic layers that cross teams and even enterprises,” he said. “The more you can simplify the modeling experience for basic users, while still accommodating the code-first users, the better you can encourage widespread usage.”
Catanzano, meanwhile, noted that improved support for integrations is an important addition to expand AtScale’s ecosystem. However, he added that such support for integrations is becoming standard, so although compelling, it is not a differentiator.
“The most significant [addition] is more integrations with some key vendors, but this is quickly becoming tables takes and more of a me-too,” he said.
While data sharing and a new IDE are among the more significant new capabilities, the impetus for adding the new features came in part from customer feedback, according to Dave Mariani, AtScale’s founder and chief technology officer.
In particular, users wanted improved semantic object sharing and version control of semantic models which they now get through AtScale’s integration with Git and by storing semantic models as code, he said.
Beyond customer feedback, the update targets a new user persona, Mariani continued. In addition to business analysts, data engineers are now a target audience.
“AtScale now allows data engineers and business analysts alike to collaborate on semantic modeling in one platform,” he said.
In addition to AtScale’s platform update, the vendor’s new Developer Community Edition introduces potential new customers to its semantic layer capabilities.
Included in the free version are use of AtScale’s semantic modeling language that fosters sharing and re-use of semantic objects by organizations that have adopted data mesh and similar decentralized data strategies; an easy-to-use interface and a GitHub repository for prebuilt semantic models that can be re-used across organizations.
AtScale typically charges on a consumption basis but does not publicize the details of its pricing.
Future plans
With AtScale’s platform update and the Developer Community Edition now in public preview, the vendor should continue to add integrations, according to the analysis.
Catanzano recommended that AtScale add new integrations with generative AI vendors.
Petrie, meanwhile, suggested that AtScale would be wise to add integrations with vector databases.
The vendor’s metadata hub provides a strong tool to help users give business context to both traditional AI as well as generative AI models. One potential use case for that could be enriching the vectorization of unstructured data, according to Petrie.
“I’ll be interested to see how companies actually use [the metadata hub] to enrich their labeling and vectorization of unstructured data objects,” he said. “AtScale [therefore] might want to focus more on integrating with vector databases or other elements of the GenAI ecosystem.”
Eric Avidon is a senior news writer for TechTarget Editorial and a journalist with more than 25 years of experience. He covers analytics and data management.
