Skip to main content

Introduction

AI-Link is a Python package that applies the power of AtScale’s Semantic Layer to data science applications, machine learning pipelines, and other challenges requiring a code-oriented approach.

Note: AI-Link is a programmatic interface to the Semantic Layer for streamlining code-first data workflows, not a replication of the data modeling experience offered through the AtScale canvas.

AI-Link permits collaboration among data professionals including data engineers, business intelligence analysts, and data scientists. It attributes business context to data products (e.g., models, engineered features, and predictions) and streamlines the consumption of these products across your organization and via the tools of your choice. Namely:

  • Singular feature definitions are used regardless of the consumer; there will be no misinterpretation of data or model outputs, even across teams in the organization.

  • Data and models are persisted to the warehouse level.

  • Business intelligence and artificial intelligence efforts are adjoined to make data explainable and relevant.

With AI-Link, you can:

  • simplify complex feature engineering by leveraging an existing library of common, business-vetted, and governed features.

  • write a DataFrame of predictions to a database, making them available in the Semantic Layer as a standalone table or part of an existing data model for analysis in BI tools.

  • embed machine learning models directly into the Semantic Layer, meaning predictions can be rendered automatically on new data and consumed with a BI tool.

Anyone with Python experience can use AI-Link to programmatically interact with the Semantic Layer.

Key Considerations​

AI-Link is not designed to replicate the entirety of the AtScale canvas experience in a Python environment. Rather, It allows programmatic interaction with AtScale data models as well as tools to facilitate common data workflows in Python environments.

There are many references to “features” in this documentation. To clarify:

  • The majority of references to “features” are meant in the machine learning sense – that is, a “feature” is basically analogous to a column in a dataset or a measure in AtScale.

  • On the other hand, features of this product will generally be referred to differently (e.g., as “capabilities” as opposed to “features”).

Additional Information​

See the following pages to learn more about AI-Link:

  • The Prerequisites page describes what’s needed to use AI-Link.
  • The Using AI-Link page provides an overview of AI-Link’s capabilities.
  • The API Reference offers a granular explanation of the AI-Link codebase.
  • The FAQs page answers common questions about AI-Link.