CONCEPT Cited by 1 source
Composable ML modules¶
Definition¶
Composable ML modules are self-contained, versioned executable units that encapsulate a single ML concern (data processing, model training, evaluation, deployment) and can be combined with other modules to build end-to-end pipelines. The key property is that modules are independently buildable, testable, and versionable — turning one-off jobs into a reusable catalog.
Design properties:
- Self-contained — each module is an independently executable unit
- Versioned — every code push produces a versioned artifact with tags (e.g.,
latestAlpha,latest) - Composable — modules combine into pipelines via orchestration, not tight coupling
- Namespace-isolated — teams own their module namespaces autonomously
- Catalog-forming — shared modules accumulate into a discoverable, reusable library
Seen In¶
- sources/2026-06-10-atlassian-architecting-scalable-ml-platforms — Atlassian ML Studio has 2,000+ reusable modules with 200k+ monthly iterations; replaced brittle one-off jobs with versioned artifacts that roll back via tag without redeploying infrastructure
(Source: sources/2026-06-10-atlassian-architecting-scalable-ml-platforms)