Operational AI WIKI

All 0-9 A B C D E F G H I J K L M N O P Q R S T U V W X Y Z

AI architecture

An AI architecture is a plan that describes which technologies are used and how data flows in order to implement an AI solution within an organization. Depending on the use case, an AI architecture may be part of a larger data architecture that defines the overall framework that sets the standards for how data systems will interact with one another, such as how data are collected, stored, arranged, integrated, and accessed for use.

The AI architecture should be designed in the planning phase of an AI project and is most commonly developed by an AI architect, data engineer, or other high-level technical experts. When an organization does not have the appropriate technical capabilities internally, the work of developing a comprehensive AI architecture is sometimes outsourced to technology partners.

A plan that describes which technologies are used and how data flows in order to implement an AI solution within an organization

Ideally, the AI architecture builds on an organization’s defined data architecture which is built upon the technology architecture, which is built upon a strategic enterprise architecture.


A well-developed architecture is a critical first step in the AI process because major data and technology decisions will be based on the plan. The plan should be customized for a clearly defined business use case so that its success can be measured against agreed-upon benchmarks. Further, the plan should be designed to adapt to future growth that aligns with business and budget projections.

The AI Infrastructure Alliance (AIIA) is an organization dedicated to bringing together the essential building blocks for the Artificial Intelligence applications of today and tomorrow.


At Molecula

Molecula is an operational AI company that enables businesses to deploy real-time analytics and AI through FeatureBase, its ultra low-latency database platform. FeatureBase is typically represented in both data and AI architectures. FeatureBase simplifies the flow and accessibility of data and is particularly beneficial with the development and productionization of models for ML and AI applications.

Molecula’s unique feature-first approach inverts the traditional AI architecture flow and allows the automated placement of feature extraction of all data at the beginning of the process. The impact of this architectural shift is that the once-manual process of making IT requests for big data projects can be automated and self-serve for development, analytics applications, and production. This reduces production delivery times from weeks or months to hours or days and makes the data reusable from project to project.

Screen Shot 2022 01 05 at 8.52.51 AM

A before and after AI architecture example with Molecula’s feature-first approach replacing the more manual approach that most MLOps practitioners use today.


Related Terms

Data architecture

Reference architecture

Software architecture

AI Ops

Data Ops


Technology plan



Learn More About AI Architecture 

BMC: Machine Learning & Big Data Blog

Towards Data Science: Fundamentals of Data Architecture

Wikipedia entry: Data architecture

Microsoft Azure Data Architecture Guide: Artificial intelligence architecture


WIKI Contribution

  • Have a new word to add?
  • Have an updated definition?
Become a contributor