Model registration and deployment
In recent project, I used Azure machine learing to run the python script and register and deploy the model in production.
I assisted the team with the development and deployment of the model. We prefer using Mlflow models for development and deployment, focusing on experimentation among group members in different locations rather than large-scale deployments.
For conventional web service deployments, I would create a Python script and run it in Azure Machine Learning (AML) to register the model. Our development team typically created a scoring script and a pipeline to deploy the model as a web service. For small-scale internal use, I deployed the MLflow model for quick deployment and easy tracking.
After creating the Mlflow model, we save and resister the model in Azure machine learning.
After registering the model, we retrieve it from the model list. We can perform real-time deployment, batch deployment, and web service deployment.
In a sample project, I completed a batch deployment, which was the most commonly used method in my previous project. After the deployment, I ran a job to score the new data, and I obtained the final predictions in a CSV file saved in the datastore.
Github link: https://github.com/kuldipm1984/modeldeployment
Portfolio 