Introduction to MLOps
Chapter 1: Foundations of MLOps
What is MLOps?
Why MLOps is Necessary for Machine Learning
MLOps vs. DevOps: Similarities and Differences
The Goals of an MLOps Strategy
Core Principles of MLOps
Common Challenges in Production Machine Learning
Chapter 2: The Machine Learning Lifecycle
Overview of the End-to-End ML Lifecycle
Data Ingestion and Preparation
Model Training and Experimentation
Model Evaluation and Validation
Model Deployment Strategies
Monitoring Models in Production
The Feedback Loop for Model Improvement
Chapter 3: Versioning in Machine Learning
The Importance of Reproducibility
Version Control for Code with Git
Introduction to Data Versioning
Techniques for Model Versioning
Managing Experiment Tracking
Hands-on Practical: Versioning a Simple ML Project
Chapter 4: Automation and CI/CD for ML
Introduction to CI/CD
Continuous Integration (CI) for ML Code
Continuous Delivery (CD) for Models
Continuous Training (CT)
Building a Basic ML Pipeline
Introduction to Orchestration Tools
Hands-on Practical: Create a Simple CI Pipeline with GitHub Actions
Chapter 5: Model Deployment and Serving
What is Model Deployment?
Containerization with Docker
Creating a Model API with Flask
Deployment Patterns: Online vs. Batch Prediction
Introduction to Model Registries
Fundamentals of Model Monitoring
Hands-on Practical: Package and Deploy a Model as a Docker Container
Common Challenges in Production Machine Learning
Was this section helpful?
- A Survey on Concept Drift Adaption, João Gama, Indrė Žliobaitė, Albert Bifet, Mykola Pechenizkiy, A. Bouchachia, 2014 ACM Computing Surveys, Vol. 46 (Association for Computing Machinery (ACM)) DOI: 10.1145/2523813 - A comprehensive review of the causes, detection, and mitigation strategies for concept drift in data streams, also relevant to data drift.
- Machine Learning Design Patterns, Valliappa Lakshmanan, Sara Robinson, Michael Munn, 2020 (O'Reilly Media) - This book presents design patterns for common ML problems, including those related to data validation, model monitoring, and handling data and concept drift.