Spatial-Omics Compute Evidence Architect — Translational Oncology

The mandate

Pathology images, tissue coordinates, molecular counts and cell or region annotations pass through separate instruments and analytical teams. Current studies retain final matrices and figures, but cannot always reproduce how a physical section became a computational coordinate system or which registration, segmentation and filtering choices shaped a claimed microenvironment. Translational comparison needs spatial lineage, not merely pipeline versioning.

The engagement deliverable is a Spatial-Omics Evidence Architecture for two tumour studies, covering specimen and consent boundary, block and section, orientation, tissue image, capture geometry, assay batch, molecular feature, registration, segmentation, cell or spot, region annotation, analytical transformation and research claim. It will preserve pathologist and scientist interpretation as explicit decisions rather than hidden labels.

Milestone one at week four supplies laboratory-to-analysis journeys, spatial failure modes and consequence-ranked gaps. Week nine concludes milestone two with coordinate, object and provenance models. By week sixteen, milestone three delivers reference studies and perturbation tests. The accepted architecture, workflow contracts, migration priorities and scientist-run qualification pack close milestone four at week twenty-two.

Acceptance requires pathology and bioinformatics owners to reconstruct twelve unseen findings from physical section to displayed region and underlying molecules; a rotated image and changed segmentation model must propagate without erasing prior analysis; and Research Quality must reperform sampled provenance. The science director signs after client teams run a section-adjacency and batch-confounding exercise unaided.

The client will provide de-identified specimen histories, consent constraints, section and staining records, instrument and assay logs, images, coordinate transforms, pipelines, annotations, analytical notebooks and supplier terms. Client engineers build references; pathologists and scientists determine biological meaning. Wet-lab execution, diagnostic use, endpoint validation, production implementation, publication authorship and platform procurement are excluded.

Why this is external work

Laboratories preserve specimens, imaging teams preserve pixels and bioinformatics teams preserve matrices, while consequential spatial transformations fall between them. Suppliers centre their own assay coordinate system. External architecture can join physical and computational geography without interpreting a tumour, authoring a scientific conclusion or favouring a proprietary analysis platform.

What you will own

• Map specimen, block, section, orientation, image, capture geometry, assay, feature, registration, segmentation, annotation and research claim.
• Define coordinate transformations and effective versions so a displayed region resolves to physical tissue and measured molecules.
• Design correction and comparison across adjacent sections, damaged tissue, repeated staining, batch effects and alternative segmentation.
• Exercise rotated image, shifted registration, lost fiducial, revised cell boundary, duplicated spot and confounded assay batch.
• Specify uncertainty where cellular resolution, tissue morphology and molecular capture do not support one precise biological region.
• Assess platforms through open spatial export, image provenance, pipeline reproducibility, compute burden, correction and supplier exit.
• Transfer spatial-lineage stewardship and perturbation testing to permanent pathology, laboratory and bioinformatics owners.

Candidate qualifications

• Led spatial-transcriptomic, multiplex-imaging or computational-pathology architecture across translational oncology research programmes.
• Connected physical specimens and sections to image coordinates, assay features, segmentation, annotations and analytical claims.
• Reconstructed findings after orientation, registration, fiducial, segmentation or batch changes altered apparent tissue geography.
• Worked with pathologists and biologists without allowing computational regions or cell labels to impersonate expert interpretation.
• Governed de-identified but sensitive specimen evidence, large image assets and reproducible compute across partner laboratories.
• Delivered supplier-neutral architecture that client scientists qualified through unseen spatial and analytical perturbations after closure.

Non-negotiables

• The named architect must lead Hyderabad workflow observation and both section-adjacency and batch-confounding acceptance exercises.
• No financial relationship may exist with spatial-assay, imaging, annotation or bioinformatics-platform suppliers evaluated.
• Pathologists and authorised scientists retain specimen, biological, diagnostic and publication decisions.
• Wet-lab delivery, diagnostic use, endpoint validation and production software implementation are expressly excluded.

1. 49 words maximum. Describe a spatial-omics finding that changed after registration or segmentation was corrected.
2. 49 words maximum. How did you link a computational region back to physical tissue without overstating cellular precision?
3. 49 words maximum. Which client artefacts are essential before testing a rotated-section scenario?