MLOps Workload Orchestrator adds support for Amazon SageMaker Model Explainability and Model Bias Monitoring
AWS Solutions has updated the MLOps Workload Orchestrator (formerly known as the AWS MLOps Framework), an AWS Solutions Implementation that streamlines the pipeline deployment process and enforces architecture best practices for machine learning (ML) model model operationalization. This solution addresses common operational pain points, including model monitoring and multi-account governance, that customers face when adopting multiple ML workflow automation tools.
AWS Solutions has updated the MLOps Workload Orchestrator (formerly known as the AWS MLOps Framework), an AWS Solutions Implementation that streamlines the pipeline deployment process and enforces architecture best practices for machine learning (ML) model model operationalization. This solution addresses common operational pain points, including model monitoring and multi-account governance, that customers face when adopting multiple ML workflow automation tools.
This update adds two new pipelines to deploy Amazon SageMaker Clarify Model Explainability and Amazon SageMaker Clarify Model Bias baseline jobs and monitoring schedules. The added pipelines help data scientist/ML engineers monitor the feature attribution drift and model bias, respectively, on a regular basis, and generate alerts if issues are detected.
This solution provides the following key features:
- Initiates a pre-configured pipeline through an API call or a Git repository
- Automatically deploys a trained model and provides an Amazon SageMaker inference endpoint
- Continuously monitors deployed machine learning models and detects any deviation in their data quality, model quality, model explainability, and/or model bias.
- Supports running your own integration tests to ensure that the deployed model meets expectations
- Allows provisioning of multiple environments to support your ML model’s life cycle
- The option to use Amazon SageMaker Model Registry to deploy versioned models
- Multi-account support for bring-your-own-model and model monitor pipelines
- Allows customers to build and register docker images for custom algorithms, to be used for model deployment on an Amazon Sagemaker endpoint.
For details about this solution, visit the solution’s AWS Solutions Implementation webpage.
Additional AWS Solutions are available on the AWS Solutions Implementation webpage, where customers can browse solutions by product category or industry to find AWS-vetted, automated, turnkey reference implementations that address specific business needs.