Google Professional Machine Learning Engineer

Every question in the Google Professional Machine Learning Engineer 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 Google Professional Machine Learning Engineer Foundation question bank, 60 questions, 4 domains. Published by CERTILUM. This is Certilum’s own measurement of its own question bank and is not an official Google Cloud statistic.