Foundation models dominate the AI conversation, but most real enterprise value still sits in structured data — churn, fraud, pricing, demand, risk — where AutoML quietly outperforms LLMs on accuracy, cost, and reliability. This 45-minute masterclass, based on the O'Reilly book *Learning AutoML*, gives technical practitioners a clear framework for when to reach for AutoML, when to reach for foundation models, and when the right answer is to combine them.

The session opens with a decision framework: where AutoML wins on tabular and time-series problems, where LLMs add leverage (feature engineering, semi-structured data, explanations), and the failure modes of each. We then run a live, end-to-end walkthrough on a real dataset — data prep, model search, leaderboard interpretation, calibration, and deployment — using AutoGluon as the working example. The final segment covers what it takes to industrialize AutoML in production: governance, monitoring, drift, retraining cadence, and the organizational patterns that separate pilots from durable systems.

Attendees will leave with a working mental model for AutoML vs. foundation models, a reusable code template, and a checklist for taking AutoML from notebook to production.

Format: Lecture with live demo. Technical literacy assumed (Python, basic ML). Duration: 45 minutes.