{"id":129602,"date":"2021-07-08T08:00:08","date_gmt":"2021-07-08T12:00:08","guid":{"rendered":"https:\/\/www.kdnuggets.com\/?p=129602"},"modified":"2021-07-08T07:25:14","modified_gmt":"2021-07-08T11:25:14","slug":"mlops-is-an-engineering-discipline","status":"publish","type":"post","link":"https:\/\/www.kdnuggets.com\/2021\/07\/mlops-engineering-discipline.html","title":{"rendered":"MLOps is an Engineering Discipline: A Beginner’s Overview"},"content":{"rendered":"
comments<\/a><\/div>\n

By Angad Gupta<\/a>, Data Science Student<\/b><\/p>

\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\"NVIDIA
\nLearn more<\/p>\n<\/a><\/div><\/div>\n

Introduction<\/b><\/h3>\n

 
\nMLOps is a combination of ML + DEV + OPS. MLOps basically helps to increase production scalability and quality of production models by increasing automation.<\/p>

\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\"NVIDIA
\nLearn more<\/p>\n<\/a><\/div><\/div>\n

MLOps is the idea of combining the\u00a0long-established practice of DevOps with the emerging field of Machine Learning. It is the creation of an automated environment for model development, model retraining, drift monitoring, automation of pipeline, quality control, and\u00a0governance of a model into a single platform.<\/p>\n

\"Figure\"
\nImage source: techinnocens<\/font><\/center>
\n <\/p>\n

An MLOps team includes the data scientists who curate datasets and design AI models and ML engineers who run those models and datasets in the automated ways.<\/p>\n

 <\/p>\n

Why MLOps is important<\/b><\/h3>\n

 
\nAn MLOps team will help you the following issues:<\/p>\n

Deployment issues:<\/b><\/p>\n

    \n
  1. Machine learning build with multiple languages\n
  2. Model deployment on development & production environments\n
  3. Troubleshooting issues raised during model deployments\n
  4. Preparedness of deployment packages with different languages\n<\/ol>\n

    Monitoring Issues:<\/b><\/p>\n

      \n
    1. Model performance monitoring\n
    2. Consistent way to monitor the models deployed across the organization\n<\/ol>\n

      Model lifecycle management issues:<\/b><\/p>\n

        \n
      1. Needing the involvement of data scientists to update the production models and maintenance activities\n
      2. Keeping track of model decay after initial deployments\n<\/ol>\n

        Model Governance:<\/b><\/p>\n

          \n
        1. Production access control\n
        2. Traceable model results\n
        3. Model audit trails\n
        4. Model upgrade approval workflows\n<\/ol>\n

           <\/p>\n

          Goals of MLOps<\/b><\/h3>\n

           
          \nThe goals of MLOps include:<\/p>\n