Usage

Deployed Resources

Once deployed, you should see the following in your AWS account:

Naming convention: <org>-<env>-<domain>-<module>-<resource>

SSM parameters: /<org>/<domain>/<module>/<resource-type>/<resource-name>/<attribute>

Resource	Deployed Name / SSM Path	Config Reference
CodePipeline Pipeline (Training)	<org>-dev-mlops-core-<project>-build /<org>/mlops/core/pipeline/training/name	training in mlops/mlops.yaml
CodePipeline Pipeline (Deploy)	<org>-dev-mlops-core-<project>-deploy /<org>/mlops/core/pipeline/deploy/name	deploy in mlops/mlops.yaml
SageMaker Model Package Group	<org>-dev-mlops-core-<project>-mpg /<org>/mlops/core/model-package-group/name	deployed by mlops/mlops.yaml (not configurable)
S3 Bucket (Artifacts)	<org>-dev-mlops-core-model-<project> /<org>/mlops/core/bucket/name	deployed by mlops/mlops.yaml (not configurable)
CodeCommit Repository (Training)	<org>-dev-mlops-core-<project>-build /<org>/mlops/core/repository/training/name	training.seedCodePath in mlops/mlops.yaml
CodeCommit Repository (Deploy)	<org>-dev-mlops-core-<project>-deploy /<org>/mlops/core/repository/deploy/name	deploy.seedCodePath in mlops/mlops.yaml

Post-Deployment Steps

1. Verify Pipeline Execution

The training pipeline starts automatically after deployment:

1. Navigate to AWS CodePipeline Console > select <org>-dev-mlops-core-<project>-build.
2. Confirm the pipeline execution is in progress or has completed successfully.
3. If the Source stage fails, verify the CodeCommit repository contains the seed code.

2. Approve the Trained Model

Once training completes, the model is registered in SageMaker Model Registry in "PendingManualApproval" status:

1. Navigate to SageMaker Console > Model Registry > select <org>-dev-mlops-core-<project>-mpg.
2. Select the latest model version.
3. Change the status to Approved to trigger the deploy pipeline.

3. Verify Endpoint Deployment

After approval, EventBridge triggers the deploy pipeline automatically:

1. Navigate to CodePipeline Console > select <org>-dev-mlops-core-<project>-deploy.
2. Confirm the pipeline completes successfully.
3. Navigate to SageMaker Console > Inference > Endpoints to confirm the endpoint is "InService".

Updating Seed Code

CloudFormation does not update existing CodeCommit repositories on subsequent deployments. To update training or deployment code:

# Clone the training repo
git clone codecommit://<org>-dev-mlops-core-<project>-build

# Make changes, commit, and push to trigger a new pipeline execution
git push