NVIDIA Certified Professional - Agentic AI

Every question in the NVIDIA Certified Professional - Agentic AI bank is measured on four signals that can be read directly from the database. Each signal is expressed as a share of the questions in the domain, and the four shares are combined with fixed weights:

• Multi-select questions (40% weight). Items with more than one correct answer. The candidate has to find all of them, which is the single hardest format in the bank.
• Scenario questions (30% weight). Questions longer than 150 characters, which describe a situation instead of a definition and therefore test reasoning rather than recall.
• Interactive questions (20% weight). Items that require reading code or a configuration snippet.
• Deep explanations (10% weight). Questions whose explanation runs past 300 characters, which signals a concept with real nuance behind it.

The combined share is multiplied by ten to produce a 1-10 score. The exam-level score is computed over the whole bank rather than averaged across domains, so a small domain cannot distort the total. Figures on this page are recomputed from the live bank on every crawl.

Data set: Certilum NVIDIA Certified Professional - Agentic AI Foundation question bank, 60 questions, 10 domains. Published by CERTILUM. This is Certilum’s own measurement of its own question bank and is not an official NVIDIA statistic.