Professional Machine Learning Engineer exam
50-60 multiple choice and multiple select questions.
- Type
- Written
- Delivery
- Both
- Duration
- 120 min
Exam sections
Architecting low‑code AI solutions
Design and build AI/ML solutions using Google‑managed services with minimal custom code. Key tasks include developing models with BigQuery ML, implementing AI solutions with ML APIs or foundation models, and training AutoML models for various data types like tabular, text, and video.
Preparation tips
Focus on Google-managed services like BigQuery ML and AutoML; understand when to use ML APIs versus custom models for different data formats.
Collaborating within and across teams to manage data and models
Focuses on data discovery, ingestion, and preprocessing across teams. Includes model prototyping with Vertex AI Workbench or Colab Enterprise, managing datasets, creating and consolidating features in Feature Store, and tracking ML experiments using Vertex AI Experiments or TensorBoard.
Preparation tips
Review data ingestion from various sources (Cloud Storage, BigQuery) and experiment tracking with Vertex AI Experiments to manage cross-functional data.
Scaling prototypes into ML models
Transform prototype work into production‑ready ML models. This encompasses choosing ML frameworks and model architectures, selecting hardware (CPU, GPU, TPU) for training, implementing distributed training, and fine‑tuning foundation models or training custom models via Vertex AI.
Preparation tips
Be prepared to select hardware (CPU/GPU/TPU) and handle custom training workflows in Vertex AI, including distributed training and hyperparameter tuning.
Serving and scaling models
Deploy trained models for batch and online inference while scaling infrastructure to meet demands. Includes managing a model registry, performing A/B testing of model versions, and tuning models for production performance, latency, memory usage, and throughput optimization.
Preparation tips
Study model serving strategies for online and batch inference using Vertex AI and Dataflow, including A/B testing and scaling with Feature Store.
Automating and orchestrating ML pipelines
Build and maintain end‑to‑end ML pipelines, including continuous retraining and metadata management. Focuses on orchestrating with Vertex AI Pipelines or Cloud Composer, implementing CI/CD with Cloud Build, and tracking model lineage and metadata via Vertex ML Metadata.
Preparation tips
Familiarize yourself with orchestrating pipelines using Vertex AI Pipelines or Kubeflow and tracking metadata for lineage and auditing.
Monitoring AI solutions
Ensure AI systems remain reliable, secure, and responsible through risk identification and continuous monitoring. Involves identifying bias, implementing Google Responsible AI practices, and monitoring for training‑serving skew, performance drift, and common errors.
Preparation tips
Learn how to monitor for training-serving skew and apply Responsible AI practices like bias detection and model explainability via Vertex AI.