Machine Learning

Large language models get the attention, but classical machine learning still runs most of the world’s production AI — fraud scoring, recommendations, forecasting, ranking. More importantly, the discipline of ML — framing problems, evaluating honestly, respecting data — is exactly the discipline you need to build reliable LLM systems.

Why an LLM engineer should still learn this

The skill that separates good AI engineers from prompt tinkerers is evaluation discipline. It comes from classical ML. The vocabulary here — precision, recall, baselines, leakage, overfitting — is what you’ll use to prove your LLM feature actually works.

In this section

Learning Paradigms Supervised, unsupervised, reinforcement, and self-supervised learning — what each needs and when to reach for it.

Core Algorithms The algorithm families worth knowing — regression, trees, boosting, clustering — plus feature engineering.

Model Evaluation Metrics, baselines, cross-validation, and the data-leakage traps that produce fake accuracy.

What you’ll be able to do

Frame a real-world problem as an ML task, choose a sensible algorithm for it, and — most importantly — evaluate the result in a way that survives contact with production.

Prerequisites

AI Fundamentals, especially How Models Learn.