Online Automated Machine Learning for Streaming Data

Create, deploy, and monitor online AutoML models in production.

Why Online AutoML with KappaML?

Traditionally, machine learning model training is done in a batch setting. Individuals or teams of data scientists select the model and its hyperparameters. Over time, this process is repeated to ensure the model's accuracy. KappaML proposes a different approach: continuously training and tuning an AutoML model on a data stream.

Adaptive: Can adapt to changes in the underlying data and overcome concept drift without retraining.
Simple: Can be deployed to production in minutes and can be easily managed and monitored.
Efficient: The training process of a model happens over time. The model does not have to be retrained, resulting in low compute power requirements.
Versatile: More and more problems can now be solved with online machine learning. KappaML offers a wide range of models and algorithms to solve classification, regression, and recommendation problems.

Latest blog posts

featured image thumbnail for post The Kappa Architecture for Online Machine Learning

The Kappa Architecture for Online Machine Learning • January 09, 2022

First, a hybrid approach called the lambda architecture was introduced. This approach involves having a batch layer, a speed (stream) layer, and a serving layer. The lambda architecture provides a good balance between speed and reliability. However, the complexity introduced by having both batch and streaming pipelines can make such an architecture hard to maintain, migrate or reorganise . Due to…

Keep reading →