The AI Engineer's Guide to Surviving the EU AI Act

Chapter 2. AI Engineering: A Proactive Compliance Catalyst

AI engineering can support companies in complying with the EU AI Act by promoting appropriate tools and practices. A key part of this is machine learning operations (MLOps), which focuses on the operationalization of ML models. Together, these disciplines provide the technical components and repeatable processes necessary for designing, developing, deploying, and maintaining AI systems in a reliable and compliant way.

Implementing engineering practices such as automation, versioning, testing, reproducibility, deployment, and monitoring helps ensure AI systems meet the EU AI Act’s requirements for high-quality, safe, and trustworthy AI. This chapter introduces some practical frameworks that you can use to design and architect compliant AI systems (although these are by no means the only options). Tools like the Machine Learning Canvas and the MLOps Stack Canvas are designed to address key requirements such as risk management, technical documentation, transparency, robustness, and post-market monitoring. We’ll also discuss CRISP-ML(Q), a machine learning process model with built-in quality assurance that emphasizes risk assessment, quality assurance, comprehensive documentation, and continuous monitoring throughout the AI lifecycle. The synergy between CRISP-ML(Q) and AI engineering enables organizations to proactively engineer compliance into the AI lifecycle, ensuring that AI systems are developed and deployed in an EU ...