Doubling Down on Production AI at Tecton

Feature stores have come a long way since we launched Tecton in 2020—and the landscape of AI continues to evolve, with new tools and new ways of working constantly emerging. Organizations need to adapt in order to be successful, and Tecton is no exception. 

Over the last 5 years, we have found that what the market needs is reliable feature infrastructure for production AI. Our experience working with sophisticated AI teams during this time is that companies and teams aiming to deploy AI applications are more successful when they use a comprehensive enterprise-grade platform with uptime guarantees, SLAs, and performance at scale. 

Tecton has always focused on production AI and is doubling down on the patterns that are bringing people the most success. As such, moving forward, Tecton is going to exclusively focus on its commercial feature platform offering and will no longer be maintaining the Feast open-source project. This will allow us to focus all of our resources on making Tecton even better for teams putting AI into production.

Despite significant resources invested into Feast over the past 2 years, we haven’t been able to keep Feast aligned with the capability set offered by Tecton. The gap is accelerating. Tecton has grown beyond the original scope of a feature store to a full feature platform, with its own data engine, and automates the building, orchestration, and maintenance of data pipelines to support real-time AI in production. The fact that Feast and Tecton do not share the same code base has made it even more difficult for us to support both as we continue to prioritize advancing state-of-the-art of data infrastructure for production AI.

What this means for Feast

Feast has been useful in helping bring the concept of a feature store to the AI community, and it has served as a valuable tool for experimentation, allowing users to explore and understand the potential of feature stores in their AI workflows. 

Feast will continue to be an open-source project available to contributors from the AI community. However, even though Tecton is stepping back from the project, we want it to be successful. We’re working with others on this project to onboard new maintainers.