A fork of PostgreSQL

for real-time machine learning

What is VidarDB?

Using SQL to run ML

Write and run machine learning algorithms in SQL. Keep data in your database and run in-place machine learning algorithms as data is collected.

Real-time prediction & online learning

Make predictions while training data. Parallelize both training and output to make up-to-date predictions based on the continuously evolving models.

Support data larger than memory

In-memory like speed without in-memory cost. Our hybrid memory/disk caching solution allows you to work with data sets larger than memory with the performance you expect from an in-memory database.

Automatic crash recovery

24x7 availability with automatic recovery. VidarDB has built in recovery from disk and failover from peers.

Team collaboration

Collaborate with teammates to access, modify and train data. All the team members share the same data repository with access control, immediate visibility after one modifies the data.

Faster data injection, querying and model learning

Real-time data ingestion and instant model learning. Purpose-built database engine serves for fast data ingestion and fresh data query, as well as speedy model learning.

Typical Examples

Linear Regression

Logistic Regression


Stay in the loop.

Sign up now to get priority notifications about VidarDB.

Supported by

AC JumpStart     Velocity