ZenML, the creator of the open source Machine Learning Operationalization (MLOps) framework for data scientists, today announced it has raised $2.7 million in its seed round. It is led by Crane Venture Partners and includes notable investors, AI-researchers, and entrepreneurs like Richard Socher, Pieter Abbeel, Jim Keller, Dirk Hoke, Nicolas Dessaigne, Carsten Thoma, and others.
ZenML’s framework is cloud and tooling agnostic and is used to create production ready Machine Learning (ML) pipelines. The company will use this funding to build out the tooling suite in the open and expand its team of ML technologists.
“ZenML is designed to give data scientists a lot more ownership over their transitions to production and empowers them to communicate with production infrastructure in a much simpler way,” said Adam Probst, co-founder, ZenML. “We are passionate about the work we’re doing with data science teams and are dedicated to building a strong, sustainable open source community.”
Any data science team working to ship their models in production is faced with a deluge of custom tooling options and processes. The space has become confusing and fragmented for data scientists and engineers alike, meaning that there is no standard way of delivering business value through ML. ZenML is a tooling and infrastructure agnostic standardization layer that allows data scientists to iterate quickly on promising ideas. On the other end of the abstraction, teams can plug and play their infrastructure and tooling needs right into their ML pipeline, with a few simple configuration changes.
The ZenML framework is available as a lightweight Python library that lets data scientists express their ML workflows as pipelines. The steps within can be defined as Python functions that handle arbitrary tasks such as preprocessing data or training a model. Having defined these pipelines, ZenML now does the heavy lifting of orchestrating the workflow on any modern or legacy infrastructure. While there are other workflow automation tools that let users define workflows as pipelines, ZenML sets itself apart by treating ML-specific artifacts like models, data drift, and feature statistics as first-class citizens. The framework then offers data scientists a path to solve complex problems such as reproducibility and versioning of data, code, and models.