Google Cloud AI study guides

Interactive study guides for Google Cloud AI and machine learning certifications, with labs, calculators, sample outputs, scenarios, flashcards and practice quizzes.

or download all guides as a .zip

Professional Machine Learning Engineer (PMLE)

Follows the exam guide dated June 1, 2026. Exam sections: low-code AI ~13%, data and models across teams ~16%, scaling prototypes ~21%, serving and scaling ~20%, pipelines ~18%, monitoring ~13%. Vertex AI is now Gemini Enterprise Agent Platform; the guides use the new names and map them to the old ones. Stars show how much exam weight a guide carries (★★★ heaviest).

Suggested study path

  1. Exam overview to see the scope and the renames.
  2. The two ML fundamentals guides: most scenario questions hinge on leakage, imbalance, overfitting and choosing the right metric.
  3. The three gen AI guides: Model Garden, Agent Platform, RAG failure modes and LLM evaluation.
  4. The six exam-section guides, in order, for the Google Cloud product choices.
  5. Retake each guide's quiz until you score above 85%, then review its gotchas list the day before the exam.

Start here

All sections ★★★Exam overview and strategyWhat's tested and where study time pays off, the 2026 product renames, how to read Google scenario questions, and a "which service?" chooser.

ML fundamentals (deep dives)

Fundamentals · 2.1, 3.1 ★★★1. Data, features and class imbalanceFraming, splits, leakage, encoding, scaling, missing values, feature selection, and every way to handle imbalanced classes, with a class-weight calculator. Fundamentals · 2.3, 3.2 ★★★2. Bias-variance and evaluation metricsLearning curves, regularization, ensembles, validation strategies, precision vs recall, ROC AUC vs PR AUC, thresholds and calibration, with two calculators.

Generative AI on Google Cloud (deep dives)

Gen AI · 1.2, 2.2, 3.2 ★★★Gemini, Model Garden and Agent PlatformModel Garden (MaaS vs self-deploy), Gemini tiers and parameters, prompting, structured output, function calling, the tuning ladder, agents with ADK and Agent Runtime, and a token cost calculator. Gen AI · 1.2, 6.2 ★★★RAG and its failure modesThe RAG pipeline, Agent Search vs RAG Engine vs Agent Retrieval, chunking, embeddings, hybrid search, reranking, 16 failure modes, retrieval vs generation metrics, and a context budget calculator. Gen AI · 2.3, 6.2 ★★★LLM and agent evaluationEvaluation datasets, computation vs model-based metrics, pointwise vs pairwise, judge biases, Agent Platform Evals, trajectory metrics, safety, CI gates, and a win-rate and kappa calculator.

Exam sections

Section 1 · ~13% ★★BigQuery ML, AutoML and AI APIsThe low-code ladder, BigQuery ML model types and options, remote Gemini models from SQL, AutoML, and prebuilt APIs. Section 2 · ~16% ★★Data, Feature Store, notebooks, experimentsPreprocessing tools, data validation, the BigQuery-backed Feature Store, PII and privacy controls, Workbench vs Colab Enterprise, and experiment tracking. Section 3 · ~21% ★★★Training, tuning and hardwareModel and product choice, custom training jobs, troubleshooting, hyperparameter tuning, CPU vs GPU vs TPU, distributed strategies, and a memory calculator. Section 4 · ~20% ★★★Serving and scalingOnline vs batch, containers, Model Registry, rollouts, private endpoints, autoscaling, LLM serving, and a capacity calculator. Section 5 · ~18% ★★★Pipelines and MLOpsAgent Platform Pipelines, Managed Airflow, Ray, validation gates, retraining policies and CI/CD/CT. Section 6 · ~13% ★★★Security, responsible AI and monitoringModel Armor, safety filters, fairness metrics, Explainable AI, skew vs drift vs concept drift, and troubleshooting order.