A paradigm change is currently occurring in the field of artificial intelligence. Initially, a straightforward but ever-expensive formula, involving more data and more computation, described the course of machine learning. We started from scratch and trained ever-larger models from random initialization on enormous, task-specific datasets. Although effective, this strategy consumed a significant amount of resources, made it challenging to enter new markets, and was fragile in conditions with limited data. This book is designed for engineers, researchers, students, and technology leaders who want not only to understand this shift but also to actively participate in it. Our goal is to provide you with both the conceptual framework and the practical expertise to harness the power of pre-trained models, saving you time and computational resources, and ultimately enabling you to solve problems that were previously out of reach.