Which phase integrates the model into production systems?

Which phase integrates the model into production systems?

• A. Monitoring
• B. Deployment
• C. Evaluation
• D. Data Understanding

Answer: (B)

The phase where the model is moved into production systems and made available for real use is deployment. This step involves packaging the trained model and setting up the serving infrastructure—such as APIs or endpoints—and connecting the model to production data pipelines and feature stores. It also covers ensuring security, scalability, and manageability, including versioning and the ability to roll back if needed. Once deployed, applications and services can call the model to generate predictions in real time or on a schedule, which is the essence of integration with production environments. Monitoring, in contrast, looks at how the model performs once it’s live, to detect drift or degradation and trigger maintenance. Evaluation happens before deployment to assess performance on holdout data, and data understanding is the initial work to grasp what the data looks like and what it can support.

The phase where the model is moved into production systems and made available for real use is deployment. This step involves packaging the trained model and setting up the serving infrastructure—such as APIs or endpoints—and connecting the model to production data pipelines and feature stores. It also covers ensuring security, scalability, and manageability, including versioning and the ability to roll back if needed. Once deployed, applications and services can call the model to generate predictions in real time or on a schedule, which is the essence of integration with production environments.

Monitoring, in contrast, looks at how the model performs once it’s live, to detect drift or degradation and trigger maintenance. Evaluation happens before deployment to assess performance on holdout data, and data understanding is the initial work to grasp what the data looks like and what it can support.