Study by topic
Work through the complete question bank by official exam domain, save questions, and revisit incorrect answers.
GooglePMLE
PMLE Practice Exam — Professional Machine Learning Engineer
Professional Machine Learning Engineer tests whether you can take an ML solution from prototype to production on Google Cloud: choosing between low-code options and custom models, preparing data, training and serving at scale, automating pipelines, and monitoring models once they are live. Practise across all six official sections with full 50-question papers and a score breakdown per section.
PMLE exam at a glance
- Questions per paper
- 50
- Time limit
- 120 minutes
- Passing score
- Not published by Google
- Question format
- Multiple choice and multiple select
- Official exam fee
- USD $200
- Certification valid for
- 2 years
- ReadyForCert access
- 60 days · $8.99 AUD
- Free sample
- 20 questions
Study, practise, review
Timed exams
Take 50-question simulations with a 120-minute timer and saved progress.
Exam reports
Review your score, every answer, and your performance across each exam domain.
What the PMLE exam covers
Architecting low-code AI solutions
Solving problems with BigQuery ML, AutoML, pre-trained AI APIs and foundation models from Model Garden.
13%Collaborating within and across teams to manage data and models
Exploring and preprocessing data, Feature Store, protecting sensitive data, notebooks, and tracking experiments and evaluation metrics.
16%Scaling prototypes into ML models
Choosing a model and product, training and tuning at scale, troubleshooting training, and picking training hardware.
21%Serving and scaling models
Batch and online prediction, serving containers, the model registry and rollout strategies, and scaling for latency and throughput.
20%Automating and orchestrating ML pipelines
End-to-end pipelines, consistent preprocessing between training and serving, retraining policies, and CI/CD for ML.
18%Monitoring AI solutions
Securing AI systems, responsible AI and model interpretability, and detecting skew and drift in production.
13%
A sample PMLE question
One question written for this page, in the same style as the questions in the bank.
Monitoring AI solutions
A fraud model's precision has fallen over several weeks, although nothing in the serving code has changed. You suspect the incoming transactions no longer look like the training data. What should you set up to detect this automatically?
- AVertex AI Model Monitoring with feature skew and drift detection on the endpoint. — Correct answer
- BMore replicas behind the prediction endpoint.
- CA larger batch size for the next training run.
- DA Cloud Monitoring alert on the endpoint's request latency.
About the PMLE exam
How many questions are on the Professional Machine Learning Engineer exam?
50–60 multiple choice and multiple select questions in two hours. ReadyForCert papers use 50 questions in 120 minutes.
What score do I need to pass the PMLE exam?
Google does not publish a passing score or your numeric result — you receive a pass or fail. Because there is no published bar, aim comfortably above a bare pass: ReadyForCert scores you against a 72% threshold so you have a concrete target.
How much does the exam cost and how long is it valid?
Registration is USD $200 plus tax where applicable, and the certification is valid for 2 years.
Are these the real Google PMLE exam questions?
No. These are practice questions written against the current official exam guide. Google's exam content is confidential and the bank is rotated, so practising against leaked questions does not survive an exam refresh.
Other certifications you can practise
Same format, same price, same free quiz.
