Tecton Partners with Feast Open Source Feature Store Founder

Tecton and Feast Collaborate on Feature Store Development

Tecton, recognized as the originator of the machine learning feature store concept, has announced a collaboration with Willem Pienaar, the creator of the open-source feature store project, Feast. This partnership culminated in the release of version 0.10 of the Feast open-source tool.

Origins of the Feature Store Concept

The foundational idea for the feature store emerged from the experiences of Tecton’s founders during their tenure as engineers at Uber. Inspired by blog posts detailing Uber’s feature store implementation, Willem Pienaar independently developed Feast as an open-source equivalent.

According to Tecton co-founder and CEO Mike Del Balso, the company initially focused on delivering a best-in-class, enterprise-level feature store. Feast, created by Pienaar, quickly became the leading standard for open-source feature stores and is now a project within the Linux Foundation.

Synergy Between Tecton and Feast

Tecton subsequently brought Pienaar onto their team, where he currently leads the open-source initiatives. While an open-source product wasn’t part of the initial plan, the close alignment between the two projects made this acquisition logical.

Pienaar highlights the inherent similarities between the two platforms, describing a symbiotic relationship. Tecton’s offering represents a comprehensive, enterprise-grade solution with expanded functionality, while Feast provides a robust and proven open-source foundation.

Understanding the Feature Store

A feature store functions as a complete machine learning management system. It encompasses the data transformation pipelines that generate feature values, the storage and management of this feature data, and the delivery of consistent data sets.

Business Benefits of Open Source Contribution

Del Balso emphasizes the strategic advantage of contributing to the open-source Feast project. This involvement exposes Tecton to a broader user base, fostering a reciprocal relationship between the commercial and open-source products during their ongoing development.

Cross-Learning and Feedback Loops

Tecton actively participates in the Feast community, gaining valuable insights from diverse use cases to enhance its own product. Conversely, feedback from enterprise customers informs improvements to the open-source project, creating a beneficial feedback loop.

Future Development

Tecton intends to maintain its role as a key contributor to Feast, dedicating a team to its continued development. The latest release, version 0.10, is now available.

• Feature Store: A centralized repository for managing and serving machine learning features.
• Feature Values: The processed data points used as inputs for machine learning models.