The Professional Machine Learning Engineer exam is 2 hours, 50 to 60 multiple choice and multiple select questions, and $200 plus tax. It is offered in English and Japanese, online or at a test centre. There are no formal prerequisites but Google recommends three or more years of industry experience, including a year designing and managing solutions on Google Cloud. The six competencies are architecting low-code AI solutions, collaborating across teams to manage data and models, scaling prototypes into ML models, serving and scaling models, automating and orchestrating ML pipelines, and monitoring AI solutions. Google says the exam does not directly test coding, but you need enough Python and SQL to read code snippets. The big change this year is the rename.
In April 2026 Google announced the Gemini Enterprise Agent Platform as the evolution of Vertex AI, and the exam was updated to match. Vertex AI Pipelines becomes Agent Platform Pipelines, Workbench becomes Agent Platform Workbench, and so on. The underlying API endpoint has not changed, so this is mostly branding, but if you are using a practice test bank from 2025 you will be studying the wrong names. Check the current exam guide before you buy anything. I rate it highly because it focuses on the parts of ML that break in production: data leakage, training-serving skew, pipeline orchestration, monitoring for drift and responsible AI checks.
Those are the skills that separate someone who can train a notebook model from someone who can keep one alive. The weakness is the vendor lens. Many questions have a technically valid open source answer and a Google-managed answer, and the exam wants the Google one. That is fine for a vendor exam but it means the credential is worth far more inside a Google Cloud shop than outside it. Pair it with something vendor-neutral, like Chip Huyen's ML systems material, if you want the knowledge to travel.