Perpetual ML

Hyperparameter tuning is the slow part of training a model, and Perpetual ML removes that step. The company says this makes training more than 100x faster, and it runs without GPUs or TPUs.

Perpetual ML is a low-code and no-code machine learning suite that runs as a native app inside modern data warehouses, so your data stays where it already lives.

It is currently being developed for Snowflake, with Databricks and other warehouses planned for later.

The suite is built around PerpetualBooster, which handles the initial training.

It also supports continual learning, so models can be updated as new data arrives.

Other features include confidence intervals using Conformal Prediction, learning from geographic data, and model monitoring.

Perpetual ML covers a range of machine learning tasks, and the models are described as explainable.

Since Snowflake is the first target, check platform support before you plan around it.