Use AutoML in Azure Databricks: Review

The notebook that AutoML spits out is the only thing of value here. I downloaded it immediately. The rest is a 45-minute guided tour where you click buttons inside a sandboxed Databricks environment that someone else set up for you. You select a cluster, pick a problem type (regression), point it at a pre-loaded dataset, choose the target variable, and click 'Start AutoML'. Then you just wait.

After a few minutes, it gives you a list of models it tried. You can click to see the notebooks for the best-performing ones. This part is neat, I guess. It shows the boilerplate for an XGBoost or LightGBM model, which I can then adapt. That's the part I saved.

The actual 'learning' part of this module is text that feels like it was copied from the product docs. There's no challenge. The quiz at the end asks you things like 'What button do you click to start AutoML?'. It felt like a test to see if I was paying attention to the marketing screenshots.

It’s free, so I can't complain too much. But don't mistake this for a real hands-on lab. You're not troubleshooting, you're not writing a line of code, and you're not even preparing data. It’s a product demo. I'd have been pretty angry if I had paid for this or used a conference voucher on it.