Self-service ML platform architecture in MLOps - Mini Project: Build & Apply

Build a Simple Self-Service ML Platform Architecture

📖 Scenario: You are part of a team building a self-service machine learning platform. This platform allows data scientists to upload datasets, configure training jobs, and deploy models without deep DevOps knowledge.To start, you will create a simple architecture representation using Python dictionaries to model components and their connections.

🎯 Goal: Build a basic data structure that represents the main components of a self-service ML platform and their relationships. Then, add configuration details, implement logic to find connected components, and finally display the architecture connections.

📋 What You'll Learn

Create a dictionary representing platform components and their connections

Add a configuration variable for maximum allowed connections

Write logic to filter components based on connection count

Print the filtered components and their connections

💡 Why This Matters

🌍 Real World

ML platforms help data scientists deploy models quickly without managing infrastructure details.

💼 Career

Understanding platform architecture and configuration is key for ML engineers and MLOps specialists to build scalable, user-friendly systems.

1

Create the platform components dictionary

Create a dictionary called ml_platform with these exact entries: 'Data Ingestion': ['Data Storage'], 'Data Storage': ['Feature Store', 'Model Training'], 'Feature Store': ['Model Training'], 'Model Training': ['Model Registry'], 'Model Registry': ['Model Deployment'], 'Model Deployment': [].

MLOps

# Create the ml_platform dictionary with components and their connections
# Your code here

Need a hint?

Think of each key as a component and the list as components it connects to.

2

Add a maximum connections configuration

Add a variable called max_connections and set it to 2 to limit the number of connections a component can have.

MLOps

ml_platform = {
    'Data Ingestion': ['Data Storage'],
    'Data Storage': ['Feature Store', 'Model Training'],
    'Feature Store': ['Model Training'],
    'Model Training': ['Model Registry'],
    'Model Registry': ['Model Deployment'],
    'Model Deployment': []
}
# Add max_connections variable below
# Your code here

Need a hint?

This variable will help filter components later.

3

Filter components by connection count

Create a dictionary called filtered_components that includes only components from ml_platform whose number of connections is less than or equal to max_connections. Use a dictionary comprehension with variables component and connections.

MLOps

ml_platform = {
    'Data Ingestion': ['Data Storage'],
    'Data Storage': ['Feature Store', 'Model Training'],
    'Feature Store': ['Model Training'],
    'Model Training': ['Model Registry'],
    'Model Registry': ['Model Deployment'],
    'Model Deployment': []
}
max_connections = 2
# Create filtered_components dictionary below
# Your code here

Need a hint?

Use len(connections) <= max_connections inside the comprehension.

4

Display the filtered platform architecture

Write a print statement to display the filtered_components dictionary.

MLOps

ml_platform = {
    'Data Ingestion': ['Data Storage'],
    'Data Storage': ['Feature Store', 'Model Training'],
    'Feature Store': ['Model Training'],
    'Model Training': ['Model Registry'],
    'Model Registry': ['Model Deployment'],
    'Model Deployment': []
}
max_connections = 2
filtered_components = {component: connections for component, connections in ml_platform.items() if len(connections) <= max_connections}
# Print the filtered_components dictionary below
# Your code here

Need a hint?

The output should show all components because all have 2 or fewer connections.