Why should you care?

AwsVerse credit risk team trains models on transaction features and deploys real-time scoring endpoints.

See it live — copy this example

Run in AWS CloudShell / local AWS CLI v2, or follow the matching steps in the AWS Console (Free Tier).

aws sagemaker create-training-job \
  --training-job-name awsverse-credit-xgb \
  --role-arn arn:aws:iam::123:role/SageMakerExecutionRole \
  --algorithm-specification TrainingImage=246618743249.dkr.ecr.us-east-1.amazonaws.com/xgboost:1.7-1,TrainingInputMode=File \
  --input-data-config '[{"ChannelName":"train","DataSource":{"S3DataSource":{"S3DataType":"S3Prefix","S3Uri":"s3://awsverse-ml/train/","S3DataDistributionType":"FullyReplicated"}}}]' \
  --output-data-config S3OutputPath=s3://awsverse-ml/output/ \
  --resource-config InstanceType=ml.m5.large,InstanceCount=1,VolumeSize=30

What happened?

• Starts SageMaker training job reading CSV from S3 with XGBoost container.
• Output model artifacts land in S3 for deployment.

Practice next

1. Upload labeled dataset to S3.
2. Create SageMaker Studio lab domain.
3. Run training job; deploy model to endpoint.
4. Enable model monitor on endpoint for drift.
5. Use Spot training for cost savings.

Remember

SageMaker = managed ML lifecycle. Train on S3 data at scale. Delete endpoints when idle.

AwsVerse credit model

Underwriting needs batch and real-time inference.

Outcome: SageMaker endpoint scores applications in 40 ms at scale.