Exercise 3: Why "No MLOps" Is a Genuinely Different Kind of Gap — Possible Solution ==================================================================== WHAT "NO NEURAL NETWORKS" AND "NO NLP TECHNIQUES" ACTUALLY MEAN ------------------------------ Per this chapter's own scope note, "no neural networks" means this course never covers "genuinely deep, layered models," deferred to nn1's own job; "no NLP-specific techniques" means it never covers "turning raw text into usable features," deferred to nlp1's own job. Both of these gaps describe a MISSING TECHNIQUE — a different kind of model, or a different kind of feature-engineering step, that this course simply never taught at all, because the topic belongs to a later, more specialized course in the same subject. WHAT "NO PRODUCTION DEPLOYMENT OR MLOPS" ACTUALLY MEANS ------------------------------ Per this chapter's own scope note, this gap covers "serving [a model] reliably at scale, monitoring for real-world performance drift over time, versioning models, and automated retraining." None of these are alternative modeling TECHNIQUES in the sense the other two gaps are — they're concerns about what happens to an already-built, already- evaluated model AFTER this course's own capstone already finishes: keeping it running reliably, watching whether its real-world performance degrades over time, and managing multiple versions of it as it gets retrained. WHY THIS IS A DIFFERENT CATEGORY OF GAP ------------------------------ "No neural networks" and "no NLP techniques" are both gaps in WHICH MODELS OR FEATURES this course teaches how to build — genuinely different modeling approaches this course simply doesn't cover, each one deferred to a specific, named future course (nn1, nlp1) that will teach the missing technique directly. "No MLOps," by contrast, isn't about a missing model type or feature-engineering technique at all — a model built with exactly the techniques this course DID teach (linear regression, logistic regression, random forests) still needs deployment, monitoring, and versioning once it's actually put to real, ongoing use. This gap exists at a different STAGE of a model's own lifecycle entirely — everything covered by this course happens before a model is ever deployed; MLOps concerns begin only after that point. WHY THIS DISTINCTION IS WORTH MAKING EXPLICIT ------------------------------ Grouping all three gaps together as "things not covered" would obscure a real, useful distinction: the first two gaps will be closed by simply continuing through this course's own subject roadmap (nn1, then nlp1); the third gap belongs to an entirely different kind of skill set (software/infrastructure engineering) that isn't necessarily even part of the Data Science & ML subject's own remaining courses at all — it's a genuinely separate discipline, adjacent to but distinct from everything this course and its planned successors actually teach. WHY THIS WORKS AS AN ANSWER ------------------------------ It explains precisely what each of the three named gaps refers to using the chapter's own scope note, and distinguishes "which techniques weren't taught" (the first two, each deferred to a specific named future course) from "what happens after a model is built" (the third, a different lifecycle stage and different skill set entirely) as the real basis for treating "no MLOps" as a genuinely different category of gap.