AtScale forms a bridge between AI and BI while establishing the foundation for simplifying AI/ML pipelines and accelerating feature engineering by data scientists.
Build a Common Data Language
The AtScale semantic layer can establish a common definition of metrics and dimensions for data science and business intelligence teams.
Eliminate inconsistency and duplicative work that happens when data science and business intelligence teams look at data differently.
A single source of governed analytics, that can be continually updated and managed as the business evolves, creates trust in data and fosters a culture of collaboration across all data consumers.
Simplify Feature Engineering
AtScale includes a powerful feature design utility that supports both code-based and visual data modeling. Data teams can collaborate closely with business users on engineering features based on raw cloud data.
Transform numerical fields to categoricals. Build custom-calculated fields. Easily create time-relative features with a full range of lags and set-backs. Feature engineering at the semantic layer is dramatically simpler than physical transformations.
Harden AI/ML Data Pipelines
AtScale serves features on demand by leveraging a powerful query virtualization platform, meaning data is not physically moved in order to deliver features to models. Once a feature is published, the data pipeline is automatically defined.
Leveraging AI-Link, features can be bi-directionally managed through Python scripts. AtScale dynamically generates queries against source data based on the feature definition. This approach radically simplifies AI/ML data pipelines while hardening against disruption caused by changes to underlying data.
Publish Model Results to Existing Dashboards
Model-generated insights, like predictions or alerts, can be published back through the semantic layer to cloud data platforms. This approach lets modeled insights inherit the same semantic structure as raw data. Data scientists can simply analyze actual vs. predicted signals and analyze model drift.
By leveraging dimensionality established by BI teams, decision makers can more confidently navigate large predicted data sets — using the same time, product, and geographical hierarchies to analyze predictions as they would actual data. Further, predictions can easily be published to the same BI platforms used by the business.