AtScale, a leading semantic layer platform for data and analytics, introduced an expanded set of product capabilities to support developer productivity and business adoption of data-centric artificial intelligence (AI). These new features make use of AtScale’s distinctive location inside the data stack and support popular cloud data warehouse and lakehouse platforms like Snowflake, Databricks, Google BigQuery, Microsoft Azure Synapse, and Amazon Redshift.

Businesses in every sector are striving to maximize the returns on their AI and data science investments. According to IDC, spending on AI/ML solutions would increase 19.6% by 2023, reaching over $500B. Despite this expenditure, according to Gartner, only 54% of AI models created will reach production, as businesses find it difficult to produce enough profit to cover the cost of operationalizing models. This gap presents a huge opportunity for solutions that can shorten and hasten the process from AI/ML initiatives to business impact.

All customers utilizing AtScale AI-Link can now take advantage of two new features included in the AtScale Enterprise semantic layer platform:

  • Semantic Predictions: Semantic predictions accelerate the business outcomes of AI investments by making it easier to work with, exchange, and employ AI-generated predictions. Predictions produced by deployed AI/ML models can be written back to cloud data platforms through AtScale. The semantic model intelligence that includes dimensional consistency and discoverability is passed on to these model-generated prediction statistics. Through AtScale, predictions can be incorporated into enhanced analytics resources for a wider variety of business users and are immediately accessible for study by business users using common BI tools.
  • Managed Features: A collection of managed features for AI/ML models can be created using AtScale, which establishes a hub of centrally controlled metrics and dimensional hierarchies. Managed features can be sourced from the library of models that are currently being kept up to date by data stewards or by specific work groups. Additionally, new features produced by AI or AutoML platforms can also develop into managed features. At any step of the creation of an ML model, AtScale-controlled features are more discoverable and simpler to work with since they inherit semantic context. To train models in AutoML or other AI platforms, managed features can now be served directly via AtScale or through a feature store like FEAST.

Gaurav Rao, Executive Vice President and General Manager of AI/ML at AtScale, said, “Despite rising investments, greater adoption of AI/ML within the modern enterprise is still hindered by complexity. The need for AI is huge, exploration is on the rise, but many businesses are still not able to use the predictive insights AI models can generate. Here at AtScale we can leverage our unique position in the data stack to streamline and simplify how the business can consume and use AI immediately, generating faster time to value from their enterprise AI investments.”