Structure
Foundry itself provides four things: the pipeline engine, orchestration, the audit trail, and compute. The stages that run inside a pipeline can come from anywhere.
Data inAPI · Edge-to-Cloud · Web
Foundrypipeline · orchestration · audit trail · computeany stage is one of these four types
Müller-BBM VAS methodthe catalog
Your methodyour systems
External AIprovider systems
Engineer reviewdecision recorded
Outputexport · report · API
What Foundry provides
- Pipeline engine — pipelines are stored definitions: named, versioned sequences of stages with explicit parameters.
- Orchestration — runs the stages in order, captures each stage's output, handles calls to stages that execute elsewhere.
- Audit trail — records every call, every parameter set, every version, per pipeline run.
- Compute — assigned per workload; one measurement or ten thousand, same method definition.
What plugs in
Every stage is one of four types. All four are recorded in the same audit trail:
- Müller-BBM VAS methods — the catalog, running inside Foundry.
- Your methods — running on your systems, under your QA.
- External AI services — running on the provider's systems.
- Engineer review — the review outcome is recorded; the judgment stays with the engineer.
Data in
Measurement data enters via API Upload, Edge-to-Cloud, or Web Upload. PAK live-acquired data uses these same paths — PAK live is a transfer protocol, not a separate entry point.
Where Foundry ends
| Foundry does | Foundry doesn't |
|---|---|
| Runs the cloud-side pipeline | The edge side — that's Edge-to-Cloud |
| Provides the Müller-BBM VAS method catalog | Your methods stay on your systems, under your control |
| Calls external AI services and records the calls | The AI services run on the provider's side, under theirs |
| Runs the engineer-review step and records the outcome | The engineering judgment — that stays with the engineer |