PMLE Certification Exam Guide and Objective Map¶
โ Back to PMLE Learning Path
Source of truth: Google's Professional Machine Learning Engineer exam guide, dated June 1, 2026.
Download the official exam guide (PDF)
Exam domains and weights¶
| Official domain | Weight |
|---|---|
| 1. Architecting low-code AI solutions | ~13% |
| 2. Collaborating within and across teams to manage data and models | ~16% |
| 3. Scaling prototypes into ML models | ~21% |
| 4. Serving and scaling models | ~20% |
| 5. Automating and orchestrating ML pipelines | ~18% |
| 6. Monitoring AI solutions | ~13% |
Weights are approximate and should be treated as the relative emphasis published in the exam guide.
Objective-by-objective mapping¶
1. Architecting low-code AI solutions (~13%)¶
1.1 Developing ML models using BigQuery ML or AutoML on Gemini Enterprise Agent Platform
- Select classification, regression, forecasting, or clustering approaches based on the business problem.
- Perform feature engineering or feature selection using BigQuery ML.
- Generate predictions with BigQuery ML.
- Train models with Agent Platform AutoML.
- Fine-tune Gemini models using BigQuery.
1.2 Building AI solutions using Google Cloud AI APIs or foundation models
- Select an appropriate model from Gemini Enterprise Agent Platform Model Garden.
- Build solutions with industry APIs such as Document AI, Vision, and Translate.
- Build and tune use-case-specific solutions with Gemini, Imagen, Veo, and models as a service in Model Garden.
- Optimize Gemini applications for cost, latency, and availability.
Course mapping: Domain 1 guide and BigQuery ML lab.
2. Collaborating within and across teams to manage data and models (~16%)¶
2.1 Exploring and preprocessing data for ML
- Organize tabular, text, and image data for experimentation, training, and serving.
- Choose preprocessing tools based on scale and complexity: BigQuery SQL, Dataflow, Apache Spark, and in-memory Python frameworks.
- Create and consolidate features in Feature Store on Gemini Enterprise Agent Platform.
- Protect data privacy and handle PII.
2.2 Model prototyping using notebooks
- Apply collaboration and security practices to notebook environments.
- Develop models in Agent Platform Workbench or Colab Enterprise using PyTorch, scikit-learn, and JAX.
- Use foundation and open-source models from Model Garden in notebook prototypes.
2.3 Tracking and running ML experiments
- Select an appropriate development and experiment environment, including Experiments on Agent Platform, Agent Platform Pipelines, or Kubeflow Pipelines.
- Evaluate predictive and generative AI solutions using appropriate metrics, including LLM-as-a-judge where suitable.
- Track and compare model artifacts, versions, and lineage, including Agent Platform ML Metadata.
Course mapping: Domain 2 guide.
3. Scaling prototypes into ML models (~21%)¶
3.1 Building models for cost, complexity, latency, and scalability
- Select a model type such as ARIMA, DNN, or LLM.
- Select an appropriate product such as Agent Platform AutoML, BigQuery ML, or Agent Platform Pipelines.
- Select a deployment strategy.
- Choose modeling techniques that meet interpretability requirements.
3.2 Training models
- Organize tabular, text, speech, image, and video training data in services such as Cloud Storage and BigQuery.
- Ingest structured and unstructured data into training pipelines.
- Train with Agent Platform custom training, Kubeflow on GKE, Agent Platform AutoML, or Tabular Workflows.
- Troubleshoot training failures and perform hyperparameter tuning.
- Fine-tune foundation models from Agent Platform and Model Garden, and know when fine-tuning is appropriate.
3.3 Choosing training hardware
- Evaluate CPU, GPU, and TPU options.
- Understand distributed training on GPUs and TPUs using data and model parallelism.
Course mapping: Domain 3 guide.
4. Serving and scaling models (~20%)¶
4.1 Serving models
- Deploy batch and online inference using Agent Platform, Model Garden, Cloud Run, or GKE as appropriate.
- Package and serve models such as PyTorch and XGBoost using prebuilt or custom containers.
- Organize and version models in Agent Platform Model Registry.
- Use rollout strategies such as A/B testing and canary deployments.
- Implement inference preprocessing and postprocessing.
4.2 Scaling online model serving
- Manage and serve features with Agent Platform Feature Store.
- Deploy to public and private endpoints.
- Choose CPU, GPU, TPU, or edge hardware as appropriate.
- Scale the serving backend based on throughput, including Agent Platform Inference and containerized serving.
- Tune models for production training and serving requirements.
Course mapping: Domain 4 guide.
5. Automating and orchestrating ML pipelines (~18%)¶
5.1 Developing end-to-end ML pipelines
- Validate data and models.
- Build and orchestrate pipelines using managed or unmanaged services, templates, or custom solutions.
- Understand Agent Platform Pipelines, Managed Service for Apache Airflow, and Ray on Agent Platform.
- Ensure consistent preprocessing between training and serving.
5.2 Automating model retraining
- Determine an appropriate retraining policy.
- Deploy models through CI/CD/CT pipelines, including Cloud Build.
Course mapping: Domain 5 guide.
6. Monitoring AI solutions (~13%)¶
6.1 Identifying risks to AI solutions
- Protect systems from data or model leaks and unintentional exploitation, including data exfiltration, malicious prompting, and sharing sensitive data with LLMs.
- Understand controls such as regular expressions, safety filters, and Model Armor.
- Align with responsible AI practices, including monitoring for bias.
- Understand model explainability on Agent Platform, including Agent Platform Inference.
6.2 Monitoring, testing, and troubleshooting AI solutions
- Configure Model Monitoring on Gemini Enterprise Agent Platform for continuous evaluation of production models.
- Identify training-serving skew, data drift, concept drift, and feature attribution drift.
- Monitor, test, and evaluate generative AI solutions.
Course mapping: Domain 6 guide.
Official Google Skills resource¶
Professional Machine Learning Engineer learning path โ 17 activities
Google Skills is a curated preparation path. The exam guide remains the authority for the objectives and weights. Check each activity's current product names and prerequisites before launching a lab.
Product terminology¶
The June 1, 2026 exam guide uses the new Gemini Enterprise Agent Platform naming. Older material may refer to Vertex AI. Use Google's Vertex AI name-change reference to map older names to current names. Do not assume every legacy feature name maps one-to-one without checking the current product documentation.
๐ฌ DevopsPilot Weekly โ Learn DevOps, Cloud & Gen AI the simple way.
๐ Subscribe here