CertNexus Certified Artificial Intelligence Practitioner (CAIP) practice questions.

A practitioner reports their model has 99% accuracy on the training set but only 60% on the held-out test set. Which phenomenon best explains this gap?

A team is building an image classifier with a small labeled dataset (5,000 images across 10 classes). They want strong performance without training from scratch. Which approach is the canonical fit?

A team is building a customer-facing chatbot that must answer from the company's internal knowledge base, not the foundation model's general training data. Which architectural pattern fits?

A team needs to explain individual model predictions to end users and to regulators. Which technique provides per-prediction feature attribution that is widely supported across model types?

A monitoring system detects that a production model's prediction distribution has shifted significantly over the past month, even though input feature distributions look unchanged. Which type of drift is this?

A team is selecting a fairness criterion for a hiring model. They want each candidate with the same true qualification level to have the same probability of being shortlisted, regardless of protected-class membership. Which criterion matches?