Databricks and CD4ML

[Dunken]

I would like to use Databricks in a CD4ML way (see also https://martinfowler.com/articles/cd4ml.html). Is this possible? I would like to develop and train models in one environment once qualified, I would like to deploy the model with the application (embed), deploy it as a service or publish it as data. Can anyone point me to resources how to get started?

[User16764241763]

Hello @Armin Galliker​ 

Thank you for posting the query.

Below blogs and documentation would be a good place to start with

• https://databricks.com/product/mlflow-model-registry
• https://databricks.com/blog/2020/06/25/announcing-mlflow-model-serving-on-databricks.html
• https://docs.databricks.com/applications/mlflow/model-registry-example.html

[Dunken]

Thank you.

This is definitively a feasible "as a service" approach. What about really hosting (embed) the service running on production somewhere else (e.g. as part of my own service)? Or is this something you would not recommend?

[Dunken]

Any thoughts on this? Actually I would like to deploy the models on devices like Raspberry Pi. Is this something you support?

[Atanu]

something below you are looking for @Armin Galliker​ ?

[Dunken]

What exactly do you mean? What's below?

[Atanu · Accepted solution]

I am sorry forgot to paste the link - https://databricks.com/notebooks/computer-vision/cv_04_%20model_deployment.html . this is the one .

[jose_gonzalez]

Hi @Armin Galliker​ ,

Did @Atanu Sarkar​ reply help you to resolve this issue? Please let us know