A team needs a custom tabular model with no ML code. What fits?
Why D is the answer
Agent Platform AutoML
AutoML trains models with little or no code.
In detail
- AutoML builds a custom tabular model with no ML code.
- Use custom training for more control.
Use case
An illustrative scenario, not a real incident.
A team with tabular data and no ML engineers trains a model with AutoML.
See it in different ways
flowchart TD
A{"Where does the team work?"} -->|"SQL in BigQuery"| B["BigQuery ML"]
A -->|"No code, managed training"| C["Agent Platform AutoML"]
A -->|"Common task, no training needed"| D["Pre-built APIs: Document AI, Vision, Translate"]stateDiagram-v2 [*] --> Task Task --> PrebuiltApi: a common task Task --> BigQueryMl: data in BigQuery, SQL team Task --> AutoML: custom model, no code Task --> CustomTraining: your own code
flowchart TD
Q{"What does the team need?"}
Q -->|"No ML code"| A0["AutoML"]
Q -->|"Own training code"| A1["Custom training"]Key point: AutoML needs little or no code.
Why the other options are wrong
- A. A hand-written training loop in JAX, run on GKE nodes
- That is ML code.
- B. A distributed TPU job with custom model parallelism
- That is ML code.
- C. A Kubeflow installation that the team manages by itself
- That is a platform to run yourself.