ML Feature Lakehouse: Building Petabyte-Scale Data Pipelines with Iceberg

The term "Feature Store" might sound like just a place to store features, but in reality, it’s a powerful system for defining, managing, and deploying large-scale data pipelines. This session will simplify feature stores by breaking down the three main types and showing how they fit into an ML ecosystem. We’ll explore how feature stores enable data scientists to build, manage, and scale their pipelines, even at petabyte levels, while handling streaming data and ensuring versioning and lineage.
Join Simba Khadder, founder and CEO of Featureform, as he cuts through the jargon and delivers practical, real-world examples. You’ll learn how feature stores can be used to build scalable data pipelines for AI/ML and get a clear roadmap for integrating them into your ML workflows.
We’ll also look under the hood to see how Featureform achieves this scale using Apache Iceberg so you leave with actionable insights to improve your ML platforms and projects.

Toward the end of the session, there will also be time for Q&A.

• Data Engineers
• Data Scientists
• Machine Learning Engineers
• AI/ML Enthusiasts

Featureform is a virtual feature store that enables data scientists to define, manage, and serve their ML model's features. It sits atop existing infrastructure, transforming it into a traditional feature store. Using Featureform, data science teams can enhance collaboration, organize experimentation, facilitate deployment, increase reliability, and ensure compliance. It allows standardized definitions of transformations, features, labels, and training sets, making them easily shareable and understandable across teams. Additionally, Featureform is designed to work with individual data scientists and large enterprise teams, providing a centralized repository for machine learning resources.
Check out our open-source Feature Store here: https://github.com/featureform/featureform

Host or Publisher Marie Owens