Cross-Machine Data Fabric
One fabric serves both machines under a shared schema, so features, models, and lineage transfer between the breeder and the burner.
One fabric, two machines
The breeder (Hyperion) and the burner are different devices — a spherical tokamak and a tandem mirror — but they share one Data Fabric. The acquisition chain, validation, normalization, feature framework, archive, feature store, vector database, and governance are common; only the coordinate module, the diagnostic mix, and the per-machine bounds differ. This is deliberate.
What is shared
- The 60+ port acquisition and sub-microsecond sampling chain.
- Validation, dead-sensor imputation, and quality scoring.
- Feature definitions with machine-specific inputs bound at configuration.
- Archive, feature store, vector database, registry, and lineage.
What differs
- Coordinate frame: flux coordinates for the breeder, mirror coordinates for the burner.
- Diagnostic weighting: magnetics-and-profiles versus density-potential-and-DEC.
- Physical bounds: 9.66 MA / 16.84 T versus 26.49 T plug / 17 T throat.
- Control problems: equilibrium and disruption versus plug density and ambipolar potential.
Why unify
A shared fabric means tooling, models, and operational knowledge transfer between machines, and lineage and governance are enforced once. A quality-scoring or imputation improvement helps both. It also lets the copilots and vector database reason across machines when that is useful, while keeping machine-scoped queries clean. The unified fabric is a design specification for the program whose breeder construction begins Q2 2027 and whose burner follows as a design and simulation study.