Overview

As of November 2022, Drifto brings 8 new AI-powered ops to your SQL interface with DriftoQL. DriftoQL’s ops range from sentiment scoring to text tagging to model training to logical inference.

​

Architecture

Drifto operates as a query engine. This means that your data and model storage is not managed by Drifto. Rather, your data is stored either in your warehouse, data lake or simple storage buckets. This enables Drifto to be a lightweight (but specialized and powerful) compute system on top of your existing data stack rather than a new system that needs a heavy integration and migration.

When you issue a query to Drifto, the parser will split the query into analytical and neural subqueries. Drifto then forwards the analytical subquery to your underlying analytical engine, such as your data warehouse, while processing the neural subquery on a specialized distributed AI fleet. When the subqueries complete, Drifto merges the subquery results into one single result.

​

Environments

Drifto is a fully-managed query engine and AI service. Drifto implements Presto’s integration protocol, so it integrates seamlessly anywhere SQL runs, such as your favorite BI tool like Tableau or Mode, or a dbt pipeline. Drifto is also directly accessible through the Trino CLI.

Currently, Drifto works with the following backend warehouses:

​

Set Up

We are currently in private beta.