AI Potluck
Back to Gap Map Infrastructure / ML orchestration

Vertex AI Pipelines

Google Cloud
closed / Overall score: n/a

Serverless managed service for running machine-learning pipelines on Google Cloud. Pipelines are defined with the Kubeflow Pipelines or TFX SDK and executed without the user managing orchestration infrastructure, with runs schedulable on a recurring basis and artifact lineage tracked in ML Metadata.

Google's documentation renamed this to Agent Platform Pipelines during 2026 and the cloud.google.com/vertex-ai/docs/pipelines path now redirects to a gemini-enterprise-agent-platform URL. The slug keeps the name the corpus already uses for Google's other Vertex products; the rename is a corpus-wide question rather than this record's. Verified 2026-09-15 via the Vertex AI Pipelines documentation.

Openness

1 high confidence
1.0
license
Proprietary
service
proprietary(Google Cloud managed service)
source
closed(no implementation is published

A managed Google Cloud service. The authoring SDKs are open - Kubeflow Pipelines and TFX, both scored elsewhere on this map - but the service that executes the pipeline is unpublished and cannot be self-hosted, which is the question this dimension asks.

  • https://cloud.google.com/vertex-ai/docs/pipelines/introduction recorded 2026-09-15

    Google's own introduction: the service 'lets you automate, monitor, and govern your machine learning (ML) systems in a serverless manner by using ML pipelines to orchestrate your ML workflows', batch-running pipelines defined with the Kubeflow Pipelines or TensorFlow Extended framework. Establishes a managed serverless service authored through open SDKs but executed by a closed one.

Adoption

not assessed

No usage figure is published for this service specifically. The vendor's platform as a whole is widely used, but nothing disclosed is specific to the orchestration product being scored, and it publishes no countable artifact - no package, repository or registry entry - so no level is assigned rather than one being inferred from the parent platform.

Capability

3 medium confidence
3.0

Band 3: the service decides when each pipeline task runs and runs it on compute it manages, without the user choosing or seeing the machine. Level with the other managed pipeline services.

Verified 2026-09-15