Google Cloud PMLE glossary of training, serving, MLOps, drift, and pipeline terms.
Use this glossary when Google Cloud Professional Machine Learning Engineer (PMLE) terms start to blur together. The goal is practical recognition, not encyclopedia coverage.
| Term | Exam meaning |
|---|---|
| Vertex AI | Google Cloud platform for ML and AI development, deployment, and operations. |
| Feature | Input variable used by a model. |
| Drift | Change in input data or behavior after deployment. |
| Endpoint | Serving surface used by applications to request predictions. |
| Experiment tracking | Recording parameters, metrics, artifacts, and results for model runs. |
| Model monitoring | Operational checks for performance, quality, drift, errors, and latency. |
| Pair | How to separate them |
|---|---|
| Problem framing and data prep vs Model development | Ask which layer the scenario is testing, then match the answer to that layer only. |
| Control vs evidence | A control changes behavior; evidence proves behavior or supports investigation. |
| Managed service vs custom build | Managed services win for lower operational effort unless the requirement needs unsupported customization. |
| Prevention vs detection | Prevention blocks or reduces a bad event; detection finds or reports that it happened. |
Do not memorize terms in isolation. For each term, write one scenario where it is the best answer, one scenario where it is a distractor, and one signal that proves it worked.