Beginner

Exam Overview & Strategy

The Google Cloud Professional Machine Learning Engineer certification validates your ability to design, build, and productionize ML models using Google Cloud technologies. This lesson covers everything you need to know about the exam before you start studying.

Exam Format at a Glance

DetailValue
Number of Questions50–60 multiple choice and multiple select
Duration120 minutes (2 hours)
Registration Fee$200 USD
Passing ScoreNot publicly disclosed (estimated ~70%)
DeliveryRemote proctored or test center (Kryterion)
LanguagesEnglish, Japanese
RecertificationEvery 2 years
Prerequisites3+ years industry experience, 1+ year designing/managing GCP solutions (recommended)
💡
Tip: There are no official prerequisites — anyone can register and take the exam. However, Google recommends at least 3 years of industry experience and 1 year of hands-on GCP experience. If you are new to GCP, complete the Google Cloud ML learning path on Coursera first.

The Six Exam Domains

The exam tests six domains. Understanding the weight of each domain helps you prioritize your study time:

📊

Domain 1: Architecting Low-Code ML Solutions (~12%)

Developing ML models using BigQuery ML, AutoML, and pre-built APIs. Knowing when to use low-code vs. custom approaches.

📌

Domain 2: Collaborating within & across Teams (~16%)

ML project scoping, data collection strategies, responsible AI, and collaboration between data scientists and engineers.

🛠

Domain 3: Scaling Prototypes into ML Models (~18%)

Building models with Vertex AI, choosing frameworks, distributed training, and hyperparameter tuning at scale.

⚡

Domain 4: Serving & Scaling Models (~18%)

Model serving infrastructure, online and batch predictions, scaling endpoints, A/B testing, and optimization.

🔗

Domain 5: Automating & Orchestrating ML Pipelines (~18%)

Vertex Pipelines, Kubeflow, CI/CD for ML, feature management, and experiment tracking.

🔍

Domain 6: Monitoring ML Solutions (~18%)

Model performance monitoring, data drift detection, feature attribution, logging, and retraining triggers.

Recommended Study Plan

Here is a 4-week study plan assuming 1–2 hours per day. Adjust based on your existing GCP experience:

WeekFocusLessonsActivities
Week 1FoundationsLessons 1–2Read exam guide, explore Vertex AI console, do introductory Qwiklabs
Week 2Data & ModelsLessons 3–4Practice with BigQuery, Dataflow, build a TF model on Vertex AI
Week 3Pipelines & ServingLessons 5–6Build a Vertex Pipeline, deploy a model endpoint, set up monitoring
Week 4Practice & ReviewLessons 7–8Take practice exam, review weak areas, re-read exam tips

Question Types You Will See

The exam uses two question formats:

  • Multiple choice: Select ONE correct answer from four options. These make up the majority of questions.
  • Multiple select: Select TWO or MORE correct answers. The question will tell you how many to choose (e.g., "Select two").

Most questions are scenario-based. You will be given a business situation and asked to choose the best GCP solution. The key is identifying the constraints (cost, latency, scale, compliance) and mapping them to the right services.

⚠
Common Trap: Many questions have two "correct-looking" answers. The exam tests whether you can pick the BEST answer given specific constraints. Look for keywords like "minimize cost," "reduce latency," "simplify operations," or "maximize accuracy" — these tell you which trade-off to prioritize.

How Scoring Works

Google does not publish the exact passing score, but based on community reports the threshold is approximately 70%. Key facts about scoring:

  • There is no penalty for wrong answers — always answer every question
  • All questions are weighted equally
  • You can flag questions and return to them later
  • Your score is available immediately after completing the exam
  • Results show pass/fail only — no domain-level breakdown

Registration & Scheduling

  1. Go to cloud.google.com/certification/machine-learning-engineer
  2. Click "Register" and sign in with your Google account
  3. Choose remote proctored or test center delivery
  4. Pay the $200 USD registration fee
  5. Schedule your exam at least 1 week in advance
  6. For remote proctoring: test your system requirements (webcam, microphone, stable internet, clean desk)
💡
Pro Tip: Schedule your exam date BEFORE you start studying. Having a fixed deadline dramatically improves focus and prevents indefinite postponement. Four weeks of focused study is enough for most candidates with existing ML and cloud experience.

Prerequisites & Background Knowledge

Before diving into the domain-specific lessons, make sure you are comfortable with these fundamentals:

✅

ML Fundamentals

  • Supervised vs. unsupervised learning
  • Classification, regression, clustering
  • Training, validation, test splits
  • Bias-variance trade-off
  • Common metrics (accuracy, F1, AUC, RMSE)
✅

GCP Fundamentals

  • Projects, IAM, service accounts
  • Cloud Storage buckets and access
  • BigQuery basics (SQL, datasets, tables)
  • Cloud Console and gcloud CLI
  • Compute Engine vs. GKE vs. Cloud Run