Data science company Domino Data Lab Inc. announced today it has raised $43 million in funding and launched a new machine learning monitoring platform.

Highland Capital Partners led the round, which Dell Technologies Capital also participated in, bringing its total amount raised to $128 million.

Domino Data Lab’s central offering is its Data Science Platform, which provides a centralized environment for business intelligence teams to develop artificial intelligence models. The platform lets users define how much hardware resources should be allocated to the environment and install the analytics tools needed for their work.

It supports a wide variety of data-crunching technologies ranging from Apache Spark to proprietary software such as SAS. The service is also compatible with R and other popular programming languages in the data science community.

The new product, Domino Model Monitor, is designed to prevent AI models from exhibiting bias or degrading over time.

The problem it’s trying to solve is that most AI models are the most accurate when they’re first created. But their accuracy degrades over time because of something called “concept drift.” The statistical properties of each AI model’s variables tend to shift over time.

People’s preferences change too, or a rival company might make a shift that upends certain assumptions. The Domino Model Monitor is meant to mitigate that drift through the use of automated model-refreshing processes.

The service works by monitoring data drift and the performance of AI models in the wild, while alerting data scientists to those that show any degradation. Domino Data said it’s not uncommon to see AI models running in production degrade in accuracy by as much as 20% over time.

In addition to the new offering, Domino Data is updating its flagship Data Science Platform with new features, including the ability to fire up Spark distributed data processing clusters for data analysis in just one click. The platform also gets new project management tools to track deliverables, progress, goals and lifecycles, and support for more Kubernetes distributions, including Microsoft Azure Kubernetes.

Image: geralt/Pixabay