VMblog recently connected with a company called ZenML.  To find out more about the company and the problems that they solve, we reached out to Adam Probst, the co-founder and CEO of the company.

VMblog:  Let's kick things off by first having you explain what ZenML does for folks.

Adam Probst:  ZenML is an open-source MLOps framework built for data scientists. It provides the right abstraction layer to bring Machine Learning (ML) models from research to production as easily as possible. Data scientists don't need to know the details behind the deployment but gain full control and ownership over the whole pipelining process. ZenML standardizes writing ML pipelines across different MLOps stacks, agnostic of cloud providers, third-party vendors, and underlying infrastructure.

VMblog:  What is the main problem that ZenML solves?  And how was this problem solved before ZenML?

Probst:  Machine Learning projects are complex or sometimes even chaotic. Data scientists and Machine Learning engineers need to use many different tools to create Machine Learning pipelines. Different phases of the process also require multiple skill sets. Currently, these roles and organizations are not fitted for frictionless Machine Learning deployment. The result is huge technical debt even before converting production-ready code. ZenML closes this gap by putting data scientists in the center, giving them a unified syntax in a familiar language.

VMblog:  What is the business value delivered?

Probst:  ZenML minimizes the total time taken for a machine learning process to create business value. ZenML helps data scientists to focus on their core expertise. For the employer, this increases motivation and results in better-allocated talent with better business results. ZenML’smodularity and interoperability of tools creates greater business benefits throughout the whole process. The focus on reproducibility makes it natively easier to maintain, hand over or collaborate on existing code. Moreover, we cache the work done along the way, saving time and money already-computed sections of pipelines can be reused across the whole organization. This warm start for pipelines reduces the needed resources across the whole value chain.

VMblog:  Why should data scientists care about this?

Probst:  Data scientists want to see their ideas flourish in production. The traditional pattern of handing off models to someone else to deploy doesn’t support this aspiration and it is discouraging to experience as a data scientist.. The creativity of practitioners is limited while serving the goal of becoming production-ready. If they don't take care of this before, their work will become meaningless as it will get cut down by the engineering team. ZenML creates an easy and clear path to start with production-ready code from day one - without any caveats. ZenML is the data scientists’ door into the Ops world. We put the data science community center stage and our tool is even fully open-source. Ultimately data scientists become better in what they do because reproducibility allows them to experiment and collaborate more easily.

VMblog:  How does ZenML differentiate itself from its competitors like the big hyperscalers (AWS, Google, Azure) and other workflow automation tools (like Airflow, Prefect, Dagster or Kubeflow)?

Probst:  There are a number of key differentiating factors including:

• ZenML is open-source + an open community
• ZenML is built for data scientists with the right layer of abstraction
• ZenML provides an easy and clear path to Machine Learning in production while not being too opinionated about which tools are used
• ZenML provides comprehensive documentation and clear use cases
• ZenML is flexible enough with many horizontal and vertical integrations to existing legacy systems or new ML tools