Add Cloud Vendor Model Deployment Templates to MLOps Model Registry

Background

the GitLab MLOps model registry will support model deployments via CI files but lacks standardized templates for major cloud vendors like Google Cloud Platform (GCP) Vertex AI and Amazon Web Services (AWS) Bedrock. Adding these templates would streamline the model deployment process and improve user experience.

Problem Statement

Users need to manually configure deployment settings for each cloud vendor, which:

• Increases the risk of configuration errors
• Requires deep knowledge of each vendor's specific requirements
• Makes it difficult to maintain consistency across deployments
• Slows down the model deployment process

Proposed Solution

Implement pre-configured templates for GCP Vertex AI and AWS Bedrock deployments that include:

• Upload pre-build docker images to the providers
• Upload raw model files to the provider and make it build the docker image
• Create an endpoint
• Deploy a model to an endpoint