Study guide

1.1 PMLE Exam Facts, Policies & the June 2026 Blueprint

Google Cloud Professional Machine Learning Engineer PMLE

Track your progress as you read

Finished sections get a check mark.

1 Chapter 1: Exam Overview & the Agent Platform Map

1.1 PMLE Exam Facts, Policies & the June 2026 Blueprint 1.2 From Vertex AI to Gemini Enterprise Agent Platform: The Product Name Map

2 Chapter 2: Low-Code ML with BigQuery ML

2.1 Choosing, Training & Predicting with BigQuery ML Models 2.2 Feature Engineering & Feature Selection in BigQuery ML 2.3 Fine-Tuning & Evaluating Gemini Models from BigQuery

3 Chapter 3: AutoML & Google Cloud AI APIs

3.1 Training Models with Agent Platform AutoML 3.2 Building with Document AI, Vision, Translation & Other AI APIs

4 Chapter 4: Foundation Models in Model Garden

4.1 Evaluating & Selecting Models in Model Garden 4.2 Building & Tuning Solutions with Gemini, Imagen, Veo & MaaS 4.3 Optimizing Gemini Applications for Cost, Latency & Availability

5 Chapter 5: Exploring, Preprocessing & Protecting Data

5.1 Organizing & Exploring Tabular, Text & Image Data 5.2 Choosing a Preprocessing Tool: BigQuery, Dataflow, Spark or Python 5.3 Creating & Consolidating Features in Agent Platform Feature Store 5.4 Protecting PII & Sensitive Data in ML Workflows

6 Chapter 6: Notebooks, Prototyping & Experiment Tracking

6.1 Workbench vs. Colab Enterprise: Secure, Collaborative Notebooks 6.2 Prototyping with PyTorch, scikit-learn, JAX & Model Garden in Notebooks 6.3 Experiment Environments, Tracking & Lineage with Experiments and ML Metadata

7 Chapter 7: Evaluating Predictive & Generative AI Solutions

7.1 Evaluation Metrics for Predictive Models 7.2 Evaluating Gen AI: Computation Metrics, Rubrics & LLM-as-a-Judge

8 Chapter 8: Choosing the Modeling Approach

8.1 Choosing the Model Type: ARIMA, Trees, DNNs & LLMs 8.2 Choosing the Product & Deployment Strategy 8.3 Modeling Techniques for Interpretability Requirements

9 Chapter 9: Training Data & Training SDKs

9.1 Organizing & Ingesting Training Data from Cloud Storage & BigQuery 9.2 Custom Training on Agent Platform: Containers, Jobs & Worker Pools 9.3 Kubeflow on GKE, AutoML & Tabular Workflows as Training Choices

10 Chapter 10: Troubleshooting, Hyperparameter Tuning & Fine-Tuning

10.1 Troubleshooting ML Training Failures 10.2 Hyperparameter Tuning on Agent Platform 10.3 Fine-Tuning Foundation Models & When to Tune

11 Chapter 11: Training Hardware & Distributed Training

11.1 Choosing CPUs, GPUs & TPUs for Training 11.2 Distributed Training: Data & Model Parallelism

12 Chapter 12: Deploying Models for Inference

12.1 Batch vs. Online Inference on Agent Platform 12.2 Serving Beyond Agent Platform Endpoints: Model Garden, Cloud Run & GKE 12.3 Prebuilt & Custom Serving Containers

13 Chapter 13: Model Versions, Rollouts & Inference Logic

13.1 Organizing & Versioning Models in Model Registry 13.2 Rollout Strategies: A/B Tests, Canaries & Traffic Splits 13.3 Inference Preprocessing & Postprocessing

14 Chapter 14: Scaling Online Serving

14.1 Serving Features Online with Feature Store 14.2 Public, Dedicated & Private Endpoints 14.3 Serving Hardware & Autoscaling for Throughput 14.4 Tuning Models for Production Training & Serving

15 Chapter 15: End-to-End ML Pipelines

15.1 Validating Data & Models in Pipelines 15.2 Agent Platform Pipelines with KFP: Components, Templates & Caching 15.3 Ensuring Consistent Preprocessing Between Training & Serving

16 Chapter 16: Orchestration with Managed Airflow & Ray

16.1 Orchestrating ML Workflows with Managed Service for Apache Airflow 16.2 Ray on Agent Platform for Distributed ML Workloads

17 Chapter 17: Automating Retraining & CI/CD/CT

17.1 Choosing a Retraining Policy 17.2 CI/CD/CT for ML with Cloud Build

18 Chapter 18: Securing AI Systems, Responsible AI & Explainability

18.1 Protecting AI Systems from Data & Model Exfiltration 18.2 Defending Gen AI Apps: Safety Filters, Model Armor & Regex Guardrails 18.3 Responsible AI: Detecting & Monitoring Bias 18.4 Model Explainability on Agent Platform

19 Chapter 19: Monitoring AI Solutions

19.1 Configuring Model Monitoring on Agent Platform 19.2 Training-Serving Skew, Data Drift, Concept Drift & Attribution Drift 19.3 Monitoring, Testing & Evaluating Gen AI Solutions in Production

1.1 PMLE Exam Facts, Policies & the June 2026 Blueprint

Key Takeaways Explain This Topic AI

The Professional Machine Learning Engineer exam has 50-60 multiple-choice and multiple-select questions in two hours and costs $200 plus applicable tax.

Google's current exam guide is dated June 1, 2026 and weights six sections: 13%, 16%, 21%, 20%, 18%, and 13%.

Scaling prototypes into ML models (~21%) is the largest section, followed by serving and scaling models (~20%).

Google reports only pass or fail; candidates who fail can view a section-level score report in the CertMetrics candidate portal.

Associate and Professional exams allow four attempts in two years, with waits of 14, 60, and 365 days after the first, second, and third failures.

Last updated: September 2026

The Google Cloud Professional Machine Learning Engineer (PMLE) certification tests whether you can build, evaluate, productionize, and optimize AI solutions on Google Cloud. That includes conventional ML (tabular models, forecasting, deep learning) and generative AI (gen AI) built on foundation models. Before studying services, lock in the facts that shape how you prepare: how long the exam is, how it is scored, what happens if you fail, and exactly which topics the current exam guide lists.

Exam at a Glance

  • Item Official detail
  • Length 2 hours
  • Format 50-60 multiple-choice and multiple-select questions
  • Registration fee $200 (plus tax where applicable)
  • Languages English, Japanese
  • Delivery Online-proctored (Pearson OnVUE) or onsite at a Pearson VUE test center
  • Prerequisites None

Recommended experience 3+ years of industry experience, including 1+ year designing and managing solutions on Google Cloud

  • Current exam guide "Certification exam guide as of June 1, 2026"
  • Validity 2 years (all Professional certifications)

You register through Google's CertMetrics candidate portal (CM Connect) and then schedule with Pearson. Google moved exam delivery from Kryterion to Pearson : the last Kryterion exams ran on February 22, 2026, and the first Pearson test date was March 2, 2026. Older guides that mention Webassessor or Kryterion are out of date. The price is the same for online and onsite delivery. Reschedule or cancel at least 24 hours before your appointment , or you forfeit the exam fee. Online sessions have strict room and ID rules. For example, your legal name must match your government-issued photo ID.

Coding expectations

The exam guide says the exam does not directly assess coding skill . However, some questions include Python or SQL snippets. With minimum proficiency in both languages, you should be able to read those snippets. In practice, you need to recognize what a BigQuery ML CREATE MODEL statement does, what a Kubeflow Pipelines component decorator means, or which SDK call deploys a model. You will not write code from scratch.

Scoring, Results, and Feedback

Google Cloud exams are designed only to decide whether you meet a minimum passing standard, so results are pass/fail . Google does not publish a numeric passing score and does not give you a scaled score.

A provisional pass or fail appears on screen as soon as you submit.

Google then reviews the exam record, including compliance with the exam terms. Allow 7-10 days for the official result.

If you fail , a detailed score report appears under Exam History in the candidate portal. It breaks the exam into sections, shows about how much of the exam came from each one, and shows where you did worse. Section-level feedback is not available for beta or renewal exams.

You do not have to "pass" each section. The outcome depends on your performance across the whole exam.

Retake and Renewal Rules

The retake policy is identical for online and onsite exams and for every exam language. Every attempt counts.

  • Event Rule
  • Attempt limit Maximum of 4 attempts in a two-year period (Associate and Professional exams)
  • After a 1st failure Wait 14 days
  • After a 2nd failure Wait 60 days
  • After a 3rd failure Wait 365 days before the 4th attempt
  • Retake fee Full registration fee each time

Professional certifications are valid for two years . For professional certifications like PMLE, you renew by retaking the exam, not through continuing education. Google's renewal table lists continuing-education renewal only for Cloud Digital Leader, Associate Cloud Engineer, Professional Cloud Architect, and Professional Data Engineer. The PMLE renewal window opens 60 days before your certification's inactive date, and passing adds two years. When you first certify, you get a 50% renewal discount code. If you are less than 30 days past your inactive date, you can still renew and keep your Series ID. You cannot retake the same exam while certified unless you are inside the renewal window.

The June 2026 Blueprint

The exam page notes that the exam was updated for the transition from Vertex AI to Gemini Enterprise Agent Platform , for updates to Google Cloud's data and analytics stack, and to prioritize Google Cloud-native solutions. The six sections and their sub-sections are:

Section Weight Sub-sections

1. Architecting low-code AI solutions ~13% 1.1 Developing ML models using BigQuery ML or AutoML; 1.2 Building AI solutions using Google Cloud AI APIs or foundational models

2. Collaborating within and across teams to manage data and models ~16% 2.1 Exploring and preprocessing data; 2.2 Model prototyping using notebooks; 2.3 Tracking and running ML experiments

3. Scaling prototypes into ML models ~21% 3.1 Building models (cost, complexity, latency, scalability); 3.2 Training models; 3.3 Choosing hardware for training

4. Serving and scaling models ~20% 4.1 Serving models; 4.2 Scaling online model serving

5. Automating and orchestrating ML pipelines ~18% 5.1 Developing end-to-end ML pipelines; 5.2 Automating model retraining

6. Monitoring AI solutions ~13% 6.1 Identifying risks to AI solutions; 6.2 Monitoring, testing, and troubleshooting AI solutions

Sections 3, 4, and 5 together make up about 59% of the exam. Still, gen AI shows up across the whole blueprint, not just one section. Examples include fine-tuning Gemini from BigQuery (1.1), LLM-as-a-judge evaluation (2.3), fine-tuning foundation models (3.2), serving from Model Garden (4.1), and Model Armor (6.1).

How Scenario Questions Are Built

Most items describe an organization, a constraint, and four plausible designs. The constraint usually decides the answer:

Least operational overhead / fastest delivery favors managed and low-code options: BigQuery ML, AutoML, pre-trained APIs, or a Gemini model called through Model Garden.

Full control (custom loss, custom architecture, specific framework version) favors custom training in containers.

Latency or throughput targets point to online endpoints, autoscaling, accelerators, and model optimization. Overnight scoring points to batch inference.

Data governance (PII, perimeter, keys) points to Sensitive Data Protection, VPC Service Controls, private endpoints, and customer-managed encryption keys.

Google Cloud-native preference: when a managed Google Cloud service meets the requirement, it usually beats self-managed open-source infrastructure.

Study Plan Mapped to This Guide

Chapters 1-4: Platform names, BigQuery ML, AutoML, AI APIs, and Model Garden (Section 1).

Chapters 5-7: Data, features, privacy, notebooks, experiments, and evaluation (Section 2).

Chapters 8-11: Model and product choice, training, tuning, and accelerators (Section 3).

Chapters 12-14: Deployment, rollouts, endpoints, and serving performance (Section 4).

Chapters 15-17: Pipelines, orchestration choices, retraining, and CI/CD/CT (Section 5).

Chapters 18-19: Security, responsible AI, explainability, and monitoring (Section 6).

Pair each chapter with hands-on time in a sandbox project and with Google's sample questions from the certification page. Then use full-length timed practice to find weak sections. Two hours for up to 60 questions gives you about two minutes per item, so flag long scenarios and come back to them.

PMLE Exam Section Weights (June 1, 2026 Exam Guide)