Tecton, the machine learning (ML) feature platform company, has announced version 0.6 of its flagship feature platform. The release introduces new capabilities that accelerate the process of building production-ready features and expands support for streaming data.

Tecton 0.6 introduces notebook-driven development to build ML features and generate training datasets. Data scientists and ML engineers can now leverage Tecton’s feature engineering framework in their core modeling workflow without ever leaving their notebook. When it comes time to productionize, feature definitions can be applied to a repo and pushed to production in a matter of minutes. This unique approach offers speed and flexibility in feature development while preserving the best practices of a GitOps workflow including “features-as-code”, version control, and CI/CD.

Tecton is also introducing its stream ingest API, which provides more flexibility in managing streaming features. Tecton now provides teams with the choice to either automate their streaming pipelines with Tecton, or to transform their streaming data outside of Tecton using the stream processing engine of their choosing. Streaming data processed outside of Tecton can now be ingested directly into the feature platform, allowing teams to standardize on a single platform to store and serve all their feature data.

Streaming features need to be processed as fast as possible to provide very fresh data for more accurate predictions. Tecton 0.6 introduces a new continuous mode for non-aggregate streaming features, allowing feature data to be processed and updated within seconds of arriving from the stream. Use cases that rely on low-latency features, like fraud detection or real-time pricing, can now make more accurate predictions.