Tecton

Startup Tecton develops a machine learning feature platform used to manage the data used to power predictive machine learning models. In March Tecton unveiled the latest version of its flagship feature platform that introduced new capabilities the company said accelerate the process of building production-ready features and expands support for streaming data. (A feature is data loaded into machine learning models to make predictions.)

The new Tecton release includes notebook-driven development to build ML features and generate training datasets. That makes it possible for data scientists and machine learning engineers to leverage the Tecton feature engineering framework within their core modeling workflow, including developing and testing production-ready features, without having to leave their notebook, the company said.

The release also includes a stream ingest API, which Tecton says provides more flexibility in managing streaming features. Developer teams can either automate their streaming pipelines with Tecton or transform streaming data outside of Tecton using a stream processing engine of their choice. Streaming data processed outside of Tecton can be ingested directly into the feature platform.

The release also provides a new continuous mode for non-aggregate streaming features, allowing feature data to be processed and updated within seconds of arriving from a stream. That’s critical for prediction cases such as fraud detection and real-time pricing that rely on low-latency features, Tecton says.