Azure AutoML & Azure DevOps: 1-Wk Implementation

Many of us are familiar with Automated Machine Learning (ML) now. Most ML teams nowadays have a Data Scientist, but they still require some sort of manual intervention in building and deploying of models.

As developers, when deploying ML models to production, we need to automate the process to track, version, audit, certify and re-use every asset in our ML model lifecycle. Essentially, we extend our DevOps CI/CD pipelines (here using Azure DevOps) to handle the end-to-end ML pipeline. Data-Core offers transparency in the model creation by introducing comprehensibility in Data preprocessing and feature importance by harnessing Azure’s capabilities like Data Guardrails under the hood.

Data-Core employs Azure MLOps to build CI/CD pipelines for the native Machine Learning models built in-house, besides Azure AutoML models, to offer its users multiple predictions to compare and choose from.

The solution is based on the following three pipelines:

• Build pipeline - Builds the code and runs a suite of tests.
• Retraining pipeline - Retrains the model as per schedule or when new data is available
• Release pipeline - Operationalizes the scoring image and promotes it safely across different environments.

After deployment, a REST endpoint is generated which can be integrated with your Applications or Edge devices.

Benefits

• As it is a one-time setup [based on the target industry], less human interaction is required on a daily basis.
• Quicker time to market.
• Consistency in subsequent iterations with low error pilferage.
• Achieving satisfactory consistent results.

Enjoy a manual intervention free, end-to-end automated MLOps from sourcing data from any data store to deploying the best model to Azure Container Service (ACI) or Azure Kubernetes Service (AKS) today.