The documentation shows that Scikit-learn, TensorFlow and XGBoost models (accessible via Google AI Platform API) are supported for use within Pega. On the equivalent page for Amazon SageMaker, the explicitly supported models are followed by this statement:

"You can also connect to an Amazon SageMaker model that uses a custom algorithm. To connect to a custom model, configure the Amazon SageMaker docker container."

My team are evaluating Amazon SageMaker vs. Google Vertex AI for possible use as cloud-based ML platforms: a key factor in our decision will be their integration capabilities with Pega. Therefore, I have the following questions regarding connection to external Google models:

1. Google Vertex AI provide pre-built Docker containers for Scikit-learn, TensorFlow and XGBoost models: does Pega only support these built-in models or can the Docker containers be configured to e.g. also contain some pre-processing of input data?
2. There appears to be support for custom algorithms when using Amazon SageMaker: is equivalent functionality not available for Google Vertex AI?
   1. E.g. for leveraging models not created in Scikit-learn, TensorFlow or XGBoost: is this possible where input/output data to Google's API endpoints have formats that are consistent with the models from the supported frameworks?
   2. If custom models aren't supported via Google, is it possible to explain why this is the case given there's support via Amazon SageMaker (i.e. is there some specific limitation with Google's endpoints)? 

Many thanks in advance for any advice.