zenml

Why we included this project

ZenML targets ML and AI engineers in company settings who have outgrown notebooks and scripts and now need real orchestration. You define pipelines in Python, and they run on whatever infrastructure you already own, whether that is local, SageMaker, Vertex, or your own cluster. The framework takes care of the unglamorous parts, like containerizing your code, logging each run's metrics and metadata, and wiring in tools such as MLflow, LangGraph, or Langfuse. Because the same code can move from a classic training pipeline to an LLM or agent workflow, teams avoid rewriting everything for each backend. For small teams, that abstraction means fast iteration in development and a workable path to production-grade deployment later.