ZenML GmbH, a startup working to simplify artificial intelligence development, today announced that it has raised $3.7 million in funding.

The capital was provided as an extension to a $2.7 million seed round the company announced in 2021. Point Nine led ZenML’s latest raise. It was joined by Crane along with current and former executives from Twilio Inc., HashiCorp Inc. and Google LLC’s Kaggle unit, which operates a popular AI website of the same name.

Building an AI model involves numerous steps. Developers have to put together a training dataset, train their neural network on that dataset and run tests to determine if the neural network meets project requirements. If it doesn’t, some or all the steps involved in the process must be repeated.

Deploying an AI model to production comes with its own set of chores. Developers have to set up production infrastructure, then monitor their neural network for performance and accuracy issues. If such issues emerge, the neural network has to be retrained.

Germany-based ZenML is working to simplify the workflow. It has developed an open-source platform, also called ZenML, that allows users to create so-called machine learning pipelines. Those are software workflows that can automatically perform the key steps involved in an AI development project.

The steps in question, such as model training and deployment, are usually carried out using separate software tools. A machine learning pipeline created with ZenML can coordinate those tools to spare AI teams the hassle of doing so manually. The result, according to ZenML, is increased developer productivity.

AI teams sometimes need to replace one of the tools in a machine learning pipeline. Team members might, for example, discover a new AI training tool that is more cost-efficient than their existing software. Such software switches historically required making extensive code changes to the machine learning pipeline being updated.

ZenML promises to simplify the task. According to the company, its platform makes it possible to change the tools in a machine learning pipeline without significantly modifying that pipeline’s code. By making it easier to switch AI tools, ZenML can reduce the risk of software teams becoming locked into a specific set of products.

The company claims its approach also simplifies the deployment phase of MLOps initiatives.

Developers typically write machine learning pipelines on their local machines and then run them on their company’s cloud infrastructure. Before a locally developed pipeline can be moved to a cloud deployment, it must be rewritten to use the AI tools installed in the latter environment. ZenML says its platform reduces the amount of custom code necessary for the task. 

ZenML pipelines are written in Python, a relatively simple programming language that is widely used for AI development. The company says developers can move their existing Python pipelines to its platform with one line of code. Also in the interest of saving time, it provides prepackaged connectors for linking its platform to popular AI development tools.

The company will use the proceeds from its latest raise to support product commercialization efforts. ZenML is rolling out a paid cloud edition of its platform, ZenML Cloud, that offers additional features not included in the open-source version. The offering provides single sign-on features, access controls and integrations with more third-party development tools. 

Image: ZenML