Hi, you can only push the data from your diagram to the data pool and use it in analytics to create charts/diagram.

AFAIK you can:

• invoke a Jupyter notebook from Celonis EMS (skills) and pass parameters
• Invoke a AF from a Jupyter notebook (webhook call for example)
• Write data from Jupyter into a Data Integration table then reload the model to make that data available to Celonis
• Read from Jupyter Process Model tables via PQL, or (not recommended) Data integration tables using SQL)

and I thinks that's all

You can save your ML output as a picture in e.g. a sharepoint. Then add a picture component to the EMS referencing the picture via url. Note that the name of the saved picture shouldn't change and that it should be overwritten by every execution of your ML code to refresh on the frontend.