Testing a trained AI model on data it hasn't seen before to ensure it generalizes well and doesn't just memorize training data. Like giving students a practice test before the final exam.
After training a fraud detection model, validation testing on new transactions ensures it can accurately detect fraud it hasn't encountered before.
Model validation is a workflow step rather than a single service. Each cloud’s ML platform supports splitting data, running evaluation jobs, tracking metrics, and monitoring post-deployment performance.