Model approval workflows in MLOps - Step-by-Step Execution

Process Flow - Model approval workflows

Model Training Complete

↓

Submit Model for Approval

↓

Review Model Performance

↓

Approve

↓

Deploy Model

↓

Model Training Complete

This flow shows how a trained model is submitted for approval, reviewed, and either approved for deployment or rejected for retraining.

Execution Sample

MLOps

submit_model()
review = review_model()
if review == 'approve':
    deploy_model()
else:
    retrain_model()

This code submits a model, reviews it, and deploys if approved; otherwise, it triggers retraining.

Process Table

Step	Action	Input/Condition	Result	Next Step
1	submit_model()	Model trained	Model submitted for approval	review_model()
2	review_model()	Model metrics evaluated	Review result: approve	deploy_model()
3	deploy_model()	Review result is approve	Model deployed to production	END
4	review_model()	Model metrics evaluated	Review result: reject	retrain_model()
5	retrain_model()	Review result is reject	Model retraining started	submit_model()

💡 Process ends when model is deployed or loops back for retraining after rejection

Status Tracker

Variable	Start	After Step 1	After Step 2	After Step 3 or 5
model_status	untrained	submitted	approved or rejected	deployed or retraining

Key Moments - 2 Insights

Why does the workflow loop back to training after rejection?

Because the review_model() returns 'reject' (see execution_table row 4), retrain_model() is called to improve the model before resubmission.

What triggers the deployment of the model?

When review_model() returns 'approve' (execution_table row 2), deploy_model() is called to put the model into production.

Visual Quiz - 3 Questions

Look at the execution_table, what is the result after step 2 if the model is approved?

A Review result: approve

B Model retraining started

C Model submitted for approval

D Model deployed to production

💡 Hint

Check the 'Result' column in execution_table row 2.

At which step does the workflow start retraining the model?

A Step 3

B Step 4

C Step 5

D Step 2

💡 Hint

Look for 'retrain_model()' in the 'Action' column of execution_table.

If the model is rejected, what is the next action after retraining?

A Deploy the model

B Submit model for approval again

C End the workflow

D Review model performance

💡 Hint

See the 'Next Step' column after retrain_model() in execution_table row 5.

Concept Snapshot

Model approval workflows:
- Train model
- Submit for approval
- Review performance
- If approved, deploy
- If rejected, retrain and resubmit
This loop ensures only good models reach production.

Full Transcript

Model approval workflows start after training a model. The model is submitted for approval where its performance is reviewed. If the review approves the model, it is deployed to production. If rejected, the model is sent back for retraining to improve. This cycle repeats until the model meets approval criteria and is deployed. The process ensures quality control before deployment.