About MLOps process with KNIME

[hhkim]

Hello.

I am comparing KNIME and MLOps open source.

I think KNIME is very useful for data preparation and model creation.
But I am not sure how to deploy and manage models with KNIME.

So I would be grateful if someone can give me an example of KNIME that can replace the functionality of the below open sources

• Mlflow: Hyperparameter/Metric/Model Logging and Management
• githup : Model source code repository and version control
• jenkins/githup actions : Automate model CI/CD deployments
• fastAPI/seldon core : Model Serving

Best regards,
hhkim

[thor_landstrom]

Hello @hhki,

So to try and answer your question, KAP does have tools that can perform some of the things you are looking for but not replace all.

• MLflow - several nodes can be used to for model metrics such as the scorer node to find out how accurate your model is. Also you can probably do this with python nodes. As I will mention later below, you can also deploy your models online to the hub where it can be managed.
• Version control (similar to git/github) is supported on the hub.
• Jenkins/Github Actions: Although not quite as extensible as these two, Knime Buisness hub offers a similar experience to where you can schedule/automate workflows running based on certain criteria or events.
• for seldon core, I believe the closest thing KAP can do to it is the same as above with deploying workflows on the hub.

I do not have all the specific details across all the tools you mentioned for what KAP can do, but I listed a couple examples of how KAP fulfills each of those functionalities.

Hope this helps,
TL