What is Vertex AI?
Vertex AI is Google Cloud's managed ML platform — the GCP counterpart to SageMaker. It covers training, tuning, model registry, endpoints, and tight integration with Google's data tools and foundation models. The concepts mirror what you already know from serving and SageMaker.
Why it matters
Plenty of ML teams run on GCP, often for its data stack and access to Google's models. Being able to move between cloud ML platforms makes you flexible, and it reinforces that managed training and serving are the same idea wherever you do it. The differences are mostly names and integrations.
What to learn
- Training jobs and pipelines on Vertex
- The model registry and endpoints
- Online versus batch prediction
- Integration with BigQuery and GCS
- Access to foundation models via Vertex
- Cost management and idle resources
- Mapping Vertex concepts to SageMaker
Common pitfall
Re-learning Vertex from scratch as though it were unrelated to other platforms. The fastest path is to map each piece to what you already know — a Vertex endpoint is a SageMaker endpoint is your FastAPI service — and focus only on the GCP-specific integrations. The serving concepts transfer directly.
Resources
Primary (free):
- Google Cloud — Vertex AI docs · docs
- Google Cloud — Vertex AI quickstarts · docs
- Google Cloud — Pricing · tool
Practice
Map the SageMaker workflow you know to Vertex equivalents, then deploy a model to a Vertex endpoint and call it for a prediction. Note one integration that is distinctly GCP, like pulling training data from BigQuery. Done when you can serve a model on a second cloud platform.
Outcomes
- Run managed training and serving on Vertex AI.
- Deploy a model to an endpoint and call it.
- Use GCP data integrations for ML.
- Map Vertex concepts to SageMaker equivalents.