Machine Learning in Production: From Models to Products

After teaching our Machine Learning in Production class (formerly “Software Engineering for AI-Enabled Systems”) four times, we stupidly made a decision we soon regretted when there were still so many chapters left: We are going to write a book with our collected material.

Below is the initial form of the complete book. A final copy will be published by MIT Press. For now, cite as: Christian Kästner. Machine Learning in Production: From Models to Products. 2022.

Table of Contents

Part 1: Machine Learning in Production: Going Beyond the Model

• Introduction
• From models to systems
• Machine learning for software engineers, in a nutshell

Part 2: Requirements Engineering

• When to use machine learning
• Setting and measuring goals
• Gathering requirements
• Planning for mistakes

Part 3: Architecture and design

• Thinking like a software architect
• Quality attributes of ML components
• Deploying a model
• Automating the ML pipeline
• Scaling the system
• Planning for operations

Part 4: Quality assurance

• Quality assurance basics
• Model quality
• Data quality
• Pipeline quality
• System quality
• Testing and experimenting in production

Part 5: Process and Teams

• Data science and software engineering process models
• Interdisciplinary teams
• Technical debt

Part 6: Responsible ML Engineering

• Responsible Engineering
• Versioning, provenance, and reproducibility
• Interpretability and explainability
• Safety
• Security and privacy
• Fairness
• Transparency and accountability

All chapters are released under Creative Commons 4.0 BY-SA license.