Google Cloud Professional Machine Learning Engineer Exam Blueprint

Quick Facts

Cost

$200 USD (plus tax where applicable)

Duration

120 minutes

Questions

50-60

Passing score

Not disclosed by Google

Languages

English, Japanese

Delivery

Online-proctored OR onsite at Pearson VUE

Validity

Refer to Google Renewal FAQs

Retake wait

Not disclosed on cert page

Skills You'll Learn

Architect low-code AI solutions using BigQuery ML, AutoML, Vertex AI APIs, Model Garden. Build, train, tune custom ML models with Vertex AI custom training, Kubeflow on GKE. Fine-tune foundation models and deliver retrieval-augmented generation (RAG) on Vertex AI Agent Builder. Choose between CPU, GPU, TPU hardware and apply distributed training patterns. Serve models at scale with batch and online inference, A/B testing, Vertex AI Feature Store. Automate end-to-end ML pipelines using Vertex AI Pipelines, Kubeflow, Cloud Composer. Operationalize MLOps with CI/CD, automated retraining, model/data lineage tracking. Monitor for training-serving skew, drift, and apply Responsible AI practices. Apply explainability with Vertex Explainable AI and model evaluation including LLM-as-judge. Earn Google's professional ML credential, demonstrating production-grade ML and GenAI capability.

Exam Logistics

Question types: multiple choice and multiple select.
Recommended programming skills: Python, SQL - minimum proficiency to interpret code snippets.
Two concurrent exam guides: CURRENT (until June 1, 2026) and NEW (effective June 1, 2026).
New guide rebrands Vertex AI as 'Gemini Enterprise Agent Platform'.

Prerequisites & Recommended Experience

Recommended industry experience: 3+ years, 1+ years designing/managing solutions on Google Cloud.
No required prior certifications.
Target role: ML Engineer building, evaluating, productionizing, optimizing AI solutions.

CURRENT exam guide - sections with weights (in force until June 1, 2026)

Section 1: Architecting low-code AI solutions ~13%

BigQuery ML (classification, regression, time-series, matrix factorization).
ML APIs and foundation models (Model Garden; Document AI, Retail; RAG via Vertex AI Agent Builder).
AutoML (tabular workflows, forecasting, debugging).

Section 2: Collaborating within and across teams to manage data and models ~14%

Data prep (Cloud Storage, BigQuery, Spanner, Cloud SQL, Spark, Dataflow, TFX, Feature Store, PII/PHI).
Model prototyping (Vertex AI Workbench, Colab Enterprise; TensorFlow, PyTorch, sklearn, JAX).
Experiments (Vertex AI Experiments, Kubeflow Pipelines, TensorBoard; GenAI evaluation).

Section 3: Scaling prototypes into ML models ~18%

Building models (framework, architecture, interpretability).
Training (Vertex AI custom training, Kubeflow on GKE, distributed training, hyperparameter tuning, fine-tuning).
Hardware (CPU/GPU/TPU/edge; Reduction Server, Horovod).

Section 4: Serving and scaling models ~20%

Serving (batch/online; Vertex AI, Dataflow, BigQuery ML; PyTorch/XGBoost; A/B testing).
Scaling online serving (Feature Store; endpoints; hardware; containerized serving).

Section 5: Automating and orchestrating ML pipelines ~22%

End-to-end pipelines (validation, preprocessing, MLflow, Cloud Build, Cloud Run; Kubeflow/Vertex AI Pipelines).
Automating retraining (policy; CI/CD with Cloud Build, Jenkins).
Metadata tracking (Vertex AI Experiments, Vertex ML Metadata; lineage).

Section 6: Monitoring AI solutions ~13%

Risks (secure AI, Responsible AI, bias, explainability on Vertex AI Prediction).
Monitoring (Vertex AI Model Monitoring; training-serving skew; feature attribution drift; performance baselines).

Official Prep Resources

Cloud Skills Boost ML Engineer learning path Current Exam Guide PDF New Exam Guide PDF (effective June 1 2026) Official sample questions

Test what you've learned

Take a free GoLearnQuiz practice test. Sign in to save your score.

Beginner Warm-up and concepts Start Intermediate Real exam style Start Advanced Tough scenarios Start

Additional Helpful Details

All section weights are approximate.
Languages: English and Japanese only.
No standalone official practice exam - sample questions form is the closest equivalent.

Google Cloud Professional Machine Learning Engineer