Google Cloud Certified: Professional Machine Learning Engineer

Google Cloud Professional Machine Learning Engineer Practice Exam: PMLE Certification Questions

Upload your Google Cloud ML notes, Vertex AI and Agent Platform docs, or study PDFs, and the AI writes unlimited Professional Machine Learning Engineer practice questions with an answer key in seconds. Built for the refreshed June 2026 outline (50 to 60 questions, 2 hours, US$200) across all six domains, including the new generative AI content.

Your study files are processed securely and deleted automatically after your practice questions are built.

Upload your Google Cloud ML material and generate your first question set

Click to upload or drag and drop

PDF, DOCX, PPTX, TXT, JPG, JPEG, PNG, HEIC, ODP, ODT, BMP, or TIFF

up to 20MB

Please wait, your quiz is being created...

Uploading...

The Google Cloud Professional Machine Learning Engineer exam has 50 to 60 multiple choice and multiple select questions, runs two hours, and costs US$200 plus tax. The exam guide was refreshed on June 1, 2026 and now covers both traditional and generative AI across six domains: scaling prototypes into ML models (~21%), serving and scaling models (~20%), automating and orchestrating ML pipelines (~18%), collaborating across teams to manage data and models (~16%), architecting low-code AI solutions (~13%), and monitoring AI solutions (~13%). It is offered in English and Japanese, is valid two years, recommends three or more years of experience, and Google does not publish a passing score.

Last updated July 2026

Two things to know before you study. First, this exam was substantially refreshed: the current guide is dated June 1, 2026 and rebuilt around the Gemini Enterprise Agent Platform, Model Garden, and generative AI evaluation. Study material written for the older Vertex AI centered outline will leave real gaps. Second, the six published weights total 101 percent, not 100. That is not an error to work around: Google prefixes each weight with a tilde to mark it approximate, so use the numbers for study allocation rather than as an exact question count.

What the Professional Machine Learning Engineer exam tests

The refreshed guide splits the exam into six domains that follow an ML solution from prototype to production monitoring. No single domain dominates, which is what makes this exam broad: you cannot pass by being excellent at modeling alone. Here is what each domain actually asks.

Domain Weight What is actually in it
Scaling prototypes into ML models~21%Moving notebook experiments into production-grade training: framework choice, distributed training, hardware selection, and fine-tuning foundation models including Gemini through BigQuery ML.
Serving and scaling models~20%Online and batch serving, endpoint scaling and cost tradeoffs, latency and throughput tuning, and serving generative models and agents in production.
Automating and orchestrating ML pipelines~18%Building and scheduling pipelines, CI/CD for ML, retraining triggers, artifact and metadata tracking, and orchestrating multi-step genAI workflows.
Collaborating across teams to manage data and models~16%Exploring and preparing data, feature management, model and dataset governance, responsible AI practices, and working across data, ML, and platform teams.
Architecting low-code AI solutions~13%Choosing prebuilt and low-code options: the Gemini Enterprise Agent Platform, Model Garden, BigQuery ML, and the pretrained AI APIs, plus knowing when a custom model is the wrong answer.
Monitoring AI solutions~13%Detecting drift and skew, monitoring model and agent quality in production, evaluating genAI output, logging, and diagnosing degradation after deployment.

Weights as published in the exam guide dated June 1, 2026. Google marks each with a tilde, and as published they total 101 percent.

Because the weights are relatively flat, breadth beats depth here. The efficient approach is to pair hands-on Google Cloud work with question drilling that forces you across all six domains rather than the two you already know. Generate practice questions from your own Google Cloud ML notes so the weaker domains get exercised instead of skipped.

Questions
50 to 60
Time
2 hours
Exam fee
$200
Validity
2 years

What changed in the June 2026 refresh

This is the single most important thing to get right, because most of the free study material online predates the change. The candidate description now explicitly covers traditional and generative AI models, and the guide is built around Google's agent tooling rather than Vertex AI alone. If your notes never mention an agent platform, they are out of date.

New and heavily weighted now

The Gemini Enterprise Agent Platform runs through the guide. Alongside it: Model Garden for model selection, fine-tuning Gemini models via BigQuery ML, prompt and context engineering, retrieval augmented generation, evaluating generative AI output, and responsible AI. These are not footnotes, they are woven across several domains.

Still tested, still classical

Traditional ML has not disappeared. Distributed training, hardware choice, feature engineering, pipeline orchestration, endpoint scaling, and drift monitoring all remain. The exam now asks you to choose between a classical model, a low-code option, and a foundation model, which is a judgment call more than a recall question.

That judgment framing is why passive reading underperforms here. A question that gives you a business requirement and four plausible Google Cloud services trains the exact decision the exam is testing, and it exposes the domains where your instinct is still guessing.

Where the ML Engineer certification fits in the Google Cloud path

Google Cloud offers one associate level certification and a set of professional ones. Most engineers pass the Associate Cloud Engineer or go straight to a professional exam based on their role. Here is how the closest professional certifications compare.

  ML Engineer Data Engineer Cloud Architect
LevelProfessionalProfessionalProfessional
FocusML and genAI systemsData pipelines and warehousingBroad solution design
Questions50 to 6050 to 6050 to 60
Time2 hours2 hours2 hours
Fee$200$200$200

Working on the data side too? Compare with the Google Cloud Professional Data Engineer practice exam, or build the platform foundation with the Professional Cloud Architect practice exam and the Associate Cloud Engineer practice exam.

How to build ML Engineer practice questions that reinforce the exam

Six flat-weighted domains and a freshly rewritten guide mean your practice has to be current and broad.

1
Upload current material
Feed in notes from the June 2026 guide and current Google Cloud documentation. Older ML Engineer material predates the agent and genAI content, so check that your source covers it before you generate questions from it.
2
Cover all six domains
Because no domain is below about 13 percent, weak spots cost real points. Generate sets per domain so monitoring and low-code architecture get the same attention as training and serving.
3
Practice service selection
Write questions that give a requirement and several plausible services where only one fits. Choosing between a custom model, BigQuery ML, Model Garden, and an agent is the judgment this exam rewards.
4
Pair with hands-on work
Google recommends at least a year on the platform for a reason. Build and deploy something real, then use questions to lock in the parts your projects never touched.

Professional Machine Learning Engineer questions, answered

How hard is the Google Cloud Professional Machine Learning Engineer exam?
It is one of the harder Google Cloud professional exams, mostly because it was rewritten. You get 50 to 60 multiple choice and multiple select questions in two hours, and the June 2026 guide shifted heavily toward generative AI: the Gemini Enterprise Agent Platform, Model Garden, fine-tuning through BigQuery ML, prompt and context engineering, and evaluating genAI solutions. Candidates who studied the older Vertex AI centered material find the agent and genAI content unfamiliar, which is where most of the difficulty now sits.
What does the exam cover?
The refreshed guide lists six domains with approximate weights: scaling prototypes into ML models around 21 percent, serving and scaling models around 20 percent, automating and orchestrating ML pipelines around 18 percent, collaborating across teams to manage data and models around 16 percent, architecting low-code AI solutions around 13 percent, and monitoring AI solutions around 13 percent. Google marks every weight as approximate. The exam now covers both traditional and generative AI models.
What is the passing score for the exam?
Google does not publish a passing score for this exam. Neither the certification page nor the exam guide states a percentage or scaled threshold, and Google reports results as pass or fail without a numeric score. Any specific passing percentage you see quoted on third-party sites is an estimate, not an official figure. Plan to be comfortable across all six domains rather than aiming at a particular number.
How many questions is the exam and how long is it?
The exam has 50 to 60 multiple choice and multiple select questions and runs two hours. Registration is US$200 plus tax, it is offered in English and Japanese, and it can be taken online with remote proctoring or at a test center. The certification is valid for two years, and as of July 2026 there are three renewal paths: retake the full exam, take a shorter renewal exam, or complete designated courses in Google Skills, and the shorter renewal exam provides no score report. There is no formal prerequisite, though Google recommends three or more years of industry experience including at least one year designing and managing solutions on Google Cloud.
Is the certification worth it?
For engineers building ML and AI systems on Google Cloud, yes. It is Google's advanced ML credential and the June 2026 refresh made it one of the few vendor certifications that tests production generative AI work, not just classical ML. That currency matters: it validates agent platform, genAI evaluation, and RAG skills that employers are actively hiring for. It also counts toward Google Cloud partner requirements, so consultancies often fund it.
Do the exam guide percentages add up to 100?
Not exactly, and that is expected. The six domain weights in the current Professional Machine Learning Engineer guide total 101 percent as published, because Google prefixes each one with a tilde to mark it approximate rather than exact. This is not a typo in the guide and it does not mean the exam is misweighted. Treat the figures as rough study allocation guidance, not a precise question count per domain.

PDFQuiz is not affiliated with, endorsed by, or sponsored by Google. Google Cloud, Vertex AI, BigQuery, and Gemini are trademarks of Google LLC. This generator builds practice questions from material you upload and is a study aid, not a substitute for hands-on Google Cloud practice or official exam preparation. Exam details change, so always confirm current details on the Google Cloud certification page before you book.

Related study tools

Building the Google Cloud path? Compare the Professional Data Engineer practice exam, the Professional Cloud Developer practice exam, and the Professional Cloud DevOps Engineer practice exam. Working on ML elsewhere? See the Databricks Machine Learning Professional practice exam. Any notes work with the certification exam generator, or start from any PDF with the PDF to practice test generator.

Build your first ML Engineer practice set

Upload your Google Cloud ML notes or Agent Platform docs and generate practice questions across all six domains with an answer key in under a minute.

Start Creating Practice Questions