Use it when
- You want a well-documented tool with excellent community support.
- You want a flexible platform tool that can run from a single workstation to a large Kubernetes cluster.
- You need powerful integrated security features.
- You want to run automated checks on deployed models.
- You want to automate machine learning model training.
- You want to track logs of production ML models deployed in Docker containers.
- You want to define triggers and criteria for tasks or workflows to highlight anomalies or inefficiencies.
- You want a rich plug-in ecosystem.
- You want a developer-friendly environment.
Watch out
- Requires adaption through plug-ins to be suitable for MLOps workflows.
- Without plug-ins, the UI can be difficult to use.
- Requires careful management of plug-ins as many of the functionalities rely on them.
Example stacks
Airflow + MLflow stack