Multi-Account Model Management Guide with Amazon SageMaker and MLflow
Amazon Web Services details model management across multiple accounts and centralized topographies through the integration of MLflow and Amazon SageMaker AI Model Registry.
Amazon Web Services extends model management across different accounts following automatic model registration, making centralized and hybrid topographies available to users.
Model Management in Multiple Accounts
Following the setup of automatic model registration, model management across accounts stands out as a natural next step. The first section examined how managed MLflow synchronizes registered models with the SageMaker AI Model Registry.
New Role Definitions and Personas
In addition to the data scientist and auditor roles introduced in the first section, cross-account topographies introduce two new roles. The administrator performs one-time cross-account setups, while the model owner manages local approvals.
Setup and Prerequisites
This guide requires the completion of the first section, possession of two AWS accounts, and the deployment of the CloudFormation stack to both accounts.
Centralized Hub and Spoke Management
For large organizations with multiple teams, combining multiple development accounts with a centralized audit account is a common approach. This structure shares the MLflow application using AWS RAM.