ZenML is an extensible open-source MLOps framework designed to create reproducible pipelines.
The framework enables data scientists to write their code as automated pipelines from day one.
It is built to encourage the iterative and experimental nature of machine learning work, but also to provide a path to an automated, production-ready software base that can be deployed on any cloud or backend service.
Users can start with a simple python function, connect multiple steps as a pipeline, and set up continuous training or inference jobs on a schedule.
“In our previous startup, we had already spent many years using machine learning in production. We noticed maximum early success came with enabling full stack data science, allowing data scientists to iterate on their workflows independently, without much engineering overhead,” said Hamza Tahir, co-founder and CTO of ZenML.
“However, the problem is, how do you expose the complexity of an exploding ML tooling landscape and the modern infrastructure stack? First we addressed this gap internally. Now, by open-sourcing ZenML, we want to enable the data science community across the entire pipeline and set new standards in MLOps,” Tahir added.