Google Cloud
Professional Machine Learning Engineer is the Google Cloud professional certification for people who build, evaluate, productionize and optimize AI solutions, using both conventional machine learning and foundational models. It validates that you can build low-code models with BigQuery ML and AutoML, choose and tune Model Garden models such as Gemini, prepare data and features, prototype in Agent Platform Workbench and Colab Enterprise, train, serve and scale models on Gemini Enterprise Agent Platform (formerly Vertex AI), automate ML pipelines with CI/CD/CT and monitor production solutions with responsible AI practices. The exam does not directly assess coding, but you should be comfortable reading Python and SQL. We follow the current official exam guide, which carries no effective date and already uses the new product names.
No registration · No credit card
Question bank updated on November 9, 2026
A BigQuery ML model trained with a TRANSFORM clause that scales and bucketizes several columns returns unexpected predictions for a group of customers. You want to see the exact preprocessed feature values that the model receives for those rows, using the statistics captured at training, without retraining or rewriting the preprocessing logic. What should you do?
Your progress
The CertifyDeck Professional Machine Learning Engineer practice test includes 302+ original exam questions written from the official Google Cloud Certified Professional Machine Learning Engineer skills outline, each with the correct answer explained and available in English and Spanish.
Practice by domain with a free online test, take a timed 120-minute mock exam weighted like the real test and review your mistakes. Start for free, no sign-up or credit card, and try the practice test before deciding.
Professional Machine Learning Engineer sample questions with answers and explanations
Official weighting per domain. Start for free, no sign-up.
Machine learning engineers, data scientists and MLOps engineers who train, deploy and monitor traditional and generative AI models on Google Cloud. Google recommends 3+ years of industry experience, including 1+ years designing and managing solutions using Google Cloud. No prerequisites.
Professional Machine Learning Engineer is the Google Cloud professional certification for people who build, evaluate, productionize and optimize AI solutions with conventional machine learning and foundational models. The exam checks that you can choose between BigQuery ML, AutoML and Gemini for a given problem, prepare data and features, train and tune models, serve and scale them on Gemini Enterprise Agent Platform (formerly Vertex AI), automate MLOps pipelines and monitor them in production with responsible AI practices. It does not directly assess coding, but it includes Python and SQL code snippets you need to be able to interpret.
The official exam guide has six sections: architecting low-code AI solutions (13%), collaborating within and across teams to manage data and models (16%), scaling prototypes into ML models (21%), serving and scaling models (20%), automating and orchestrating ML pipelines (18%) and monitoring AI solutions (13%).
Low-code AI, data and experimentation: classification, regression, forecasting and clustering models with BigQuery ML or Agent Platform AutoML, feature engineering and predictions in BigQuery ML, and tuning Gemini from BigQuery; choosing models in Model Garden (Gemini, Veo and models as a service), APIs such as Document AI, Vision and Translation, and optimizing for cost, latency and availability. For data: the right preprocessing tool for the scale (BigQuery, Dataflow, Apache Spark or in-memory Python), Feature Store and handling personal data. And for experimentation: secure notebooks in Agent Platform Workbench and Colab Enterprise with PyTorch, scikit-learn or JAX, Experiments, ML Metadata, evaluation metrics and LLM-as-a-judge.
Training, serving, MLOps and monitoring: choosing the model type (ARIMA, DNN or LLM) and the product, interpretable techniques, custom training, Kubeflow on GKE and Tabular Workflows, hyperparameter tuning, training failures, fine-tuning foundational models, and distributed training on GPUs and TPUs with data and model parallelism. For serving: batch and online inference with Agent Platform, Cloud Run or GKE, prebuilt and custom containers, Model Registry, canary deployments and A/B tests, public and private endpoints, and scaling the serving backend. For MLOps: Agent Platform Pipelines, Managed Service for Apache Airflow (formerly Cloud Composer) and Ray, data and model validation, consistent preprocessing between training and serving, retraining policies and CI/CD/CT with Cloud Build. And for monitoring: Model Armor and safety filters, bias and explainability, Model Monitoring, data drift, concept drift and feature attribution drift, and evaluating generative AI solutions.
Google recommends 3+ years of industry experience, including 1+ years designing and managing solutions using Google Cloud. With that background, six to eight weeks are usually enough: read the official exam guide, follow the Machine Learning Engineer learning path in Google Skills, and build in your own project a BigQuery ML model, a custom training job with hyperparameter tuning, an endpoint with a canary rollout, an Agent Platform Pipelines run and a Model Monitoring setup. Spend more time on scaling prototypes into ML models and on serving and scaling models, which together make up 41% of the exam.
Start free to measure your level and practice by section. If you come from the data side, see [Professional Data Engineer](/en/certifications/professional-data-engineer); for a business view of generative AI on Google Cloud, [Generative AI Leader](/en/certifications/generative-ai-leader); and if you compare clouds, [MLA-C02](/en/certifications/mla-c02) covers machine learning engineering on AWS.
Official source: Official Professional Machine Learning Engineer exam guide
CertifyDeck is an independent study platform and is not affiliated with, endorsed by, or sponsored by Google Cloud. Professional Machine Learning Engineer and its related marks belong to their respective owner.
It lasts 2 hours and has 50 to 60 multiple choice questions. You can take it online with remote proctoring or at a test center. Our mock exam uses 50 questions in 120 minutes.
Google does not publish the passing score: you get a pass or fail result.
200 USD plus tax where applicable. CertifyDeck does not sell the exam, only the preparation: you can start free and full access is a one-time payment.
English and Japanese. On CertifyDeck every question and explanation is available in English and Spanish.
There are no prerequisites. Google recommends 3+ years of industry experience, including 1+ years designing and managing solutions using Google Cloud, and at least basic Python and SQL to interpret questions with code snippets.
Two years. Google also lets you renew it in Google Skills by completing the designated courses or skill badges in the renewal learning path during the last year of the certification.
Google has already updated the exam to reflect the transition from Vertex AI to Gemini Enterprise Agent Platform and the renamed products in its data and analytics stack (for example, Cloud Composer became Managed Service for Apache Airflow), and it now prioritizes Google Cloud native solutions. No further change has been announced. Our bank follows the current official exam guide with the new names, citing the old one where it helps you recognize it.
No. They are original questions written from the official exam guide and the public Google Cloud documentation.
Yes. You can start for free with no sign-up or card: Professional Machine Learning Engineer practice questions with explanations, a free online test by domain and published sample questions with answers. Full access to the whole bank and the timed mock exam is a one-off payment.
Start for free, no sign-up or credit card.
Start free, no card
Start free, no card