How do you introduce machine learning to someone just starting out? That was my challenge with a Chemical Engineering class, using real data from a project the students were building themselves. The guest lecture was part of an interdisciplinary collaboration: four lab groups investigating organic waste reuse processes, while the machine learning side analysed patterns from a community survey on waste disposal. The plan was to work through three core concepts together — clustering, Random Forest, and feature importance. I chose those deliberately. Clustering because it needs no labels and the result is immediately interpretable: here are the groups that exist in your data, do they match your intuition? Random Forest because it is forgiving of messy inputs and does not require anyone to understand gradient descent first. Feature importance because it answers the question people actually arrive with, which is not "can you predict this" but "what matters here." The thing I find works with beginners is starting from data they collected. Abstract datasets teach the mechanics but not the judgement. When it is your own survey, you already have opinions about what the answer should be — and the moment the model disagrees with you is the moment the concept lands. That happened here, in fact. One of the models had no predictive capacity at all. That turned out to be the more useful result, because it said the behaviour was driven by factors the survey never measured.