Responsible AI practices in MLOps - Mini Project: Build & Apply

Responsible AI Practices with MLOps

📖 Scenario: You are working as a machine learning engineer in a team that builds AI models. Your team wants to ensure the AI models are responsible and fair before deployment. You will create a simple project to check model fairness and document ethical considerations.

🎯 Goal: Build a small Python script that stores model predictions and true labels, sets a fairness threshold, calculates fairness metrics, and prints a fairness report. This simulates responsible AI checks in an MLOps pipeline.

📋 What You'll Learn

Create a dictionary with exact model predictions and true labels

Add a fairness threshold variable

Calculate fairness metric using a for loop

Print the fairness report with exact formatting

💡 Why This Matters

🌍 Real World

Responsible AI practices help ensure machine learning models are fair and ethical before deployment. This reduces harm and builds trust.

💼 Career

MLOps engineers and data scientists use these checks to monitor models continuously and meet ethical standards required by companies and regulators.

1

Create model predictions and true labels dictionary

Create a dictionary called model_results with these exact entries: 'predictions': [1, 0, 1, 1, 0] and 'true_labels': [1, 0, 0, 1, 0].

MLOps

# Create the model_results dictionary with predictions and true_labels
# Your code here

Need a hint?

Use a dictionary with keys 'predictions' and 'true_labels' and assign the exact lists as values.

2

Add fairness threshold variable

Add a variable called fairness_threshold and set it to 0.8.

MLOps

model_results = {'predictions': [1, 0, 1, 1, 0], 'true_labels': [1, 0, 0, 1, 0]}
# Set the fairness_threshold variable to 0.8
# Your code here

Need a hint?

Just create a variable named fairness_threshold and assign 0.8 to it.

3

Calculate fairness metric

Use a for loop with variable i to iterate over the indexes of model_results['predictions']. Calculate the accuracy by counting how many predictions match true labels. Store the accuracy as a float in a variable called accuracy.

MLOps

model_results = {'predictions': [1, 0, 1, 1, 0], 'true_labels': [1, 0, 0, 1, 0]}
fairness_threshold = 0.8
# Calculate accuracy using a for loop and store in accuracy
# Your code here

Need a hint?

Count matches between predictions and true_labels using a for loop over indexes, then divide by total count.

4

Print fairness report

Write a print statement to display the text: "Model fairness check: Accuracy = {accuracy}, Threshold = {fairness_threshold}" using an f-string with the variables accuracy and fairness_threshold.

MLOps

model_results = {'predictions': [1, 0, 1, 1, 0], 'true_labels': [1, 0, 0, 1, 0]}
fairness_threshold = 0.8
correct = 0
for i in range(len(model_results['predictions'])):
    if model_results['predictions'][i] == model_results['true_labels'][i]:
        correct += 1
accuracy = correct / len(model_results['predictions'])
# Print the fairness report using an f-string
# Your code here

Need a hint?

Use print(f"Model fairness check: Accuracy = {accuracy}, Threshold = {fairness_threshold}") to show the result.