Unlike traditional software design, ML system design is inherently ambiguous. There is no single "correct" answer to building a YouTube recommendation engine or a fraud detection pipeline; the answer depends on latency requirements, data volume, and business metrics. Xu’s book succeeds because it provides a .