SpatialX Orca: A Gigapixel Cancer-AI Platform

A build case study of Orca, an end-to-end platform for AI cancer annotation, training, and inference for pathologists. How a tiny team shipped a gigapixel whole-slide-image platform: the product problem, a generic JSON-driven frontend, an adaptive gigapixel canvas and its patent, a storage-first workflow engine where behavior is data, infrastructure as code, and compliance-by-architecture for PHI.

1. 01Why SpatialX: Reach the PathologistA build case study of Orca, SpatialX's end-to-end platform for AI cancer annotation, training, and inference (diagnosis) for pathologists. • Day 1 sets the stakes: cancer diagnosis runs on gigapixel images, and the best AI never reaches the pathologist.8 sections
2. 02Ship an AI Product with a Tiny TeamA build case study of Orca, SpatialX's gigapixel cancer-AI platform for pathologists. • Day 2: the constraints — a two-to-three-person team on a pre-seed budget — and the strategy that made shipping possible: make the easy parts a template a junior can copy-paste, so the seniors do the hard model and backend work.8 sections
3. 03The Orca Platform: One Screen, One DoorA build case study of Orca, SpatialX's gigapixel cancer-AI platform. • Day 3 draws the whole system on one page: a pathologist talks only to the frontend, everything funnels through a single API Gateway into one backend core, and three subsystems — Frontend, Backend, and the AI Services — do the rest.8 sections
4. 04The Frontend: Render Anything From JSONA build case study of Orca, SpatialX's gigapixel cancer-AI platform. • Day 4 opens the Frontend box from Day 3. Its whole job is to be a faithful, generic renderer: the backend describes a page as a list of section objects, and the frontend draws each one by type — so a new screen is a JSON edit, not new code.9 sections
5. 05The Canvas: Gigapixel on a TabletA build case study of Orca, SpatialX's gigapixel cancer-AI platform. • Day 5 is the hardest thing the frontend has to survive: showing a whole-slide image the size of a gigapixel photo, with hundreds of thousands of marks on it, smoothly, on a tablet, over a hospital network that drops.10 sections
6. 06The Patent: Adaptive RenderingA build case study of Orca, SpatialX's gigapixel cancer-AI platform. • Day 5 built the canvas and waved at the "how much detail" decision. Day 6 makes it precise: the adaptive-rendering method SpatialX filed as patent P91594EP — show only what fits the device.9 sections
7. 07The Backend: one secure coreA build case study of Orca, SpatialX's gigapixel cancer-AI platform. • The backend is one deliberately boring core, with tenant isolation built into the plumbing so a developer cannot forget it.11 sections
8. 08AI Services & Workflow Engine: behavior as dataA build case study of Orca, SpatialX's gigapixel cancer-AI platform. • The workflow engine makes behavior data: every job is an ordered list of steps stored in the database, run one storage-first loop at a time.13 sections
9. 09Infrastructure as CodeA build case study of Orca, SpatialX's gigapixel cancer-AI platform. This day: the whole platform described in code — all YAML, nothing provisioned by hand.8 sections
10. 10Compliance and HIPAA: Isolate PHIThe optional, detachable compliance chapter of the Orca case study. Compliance by architecture — a program in mid-flight toward HIPAA and ISO 15189, honestly not yet certified.9 sections
11. 11Tradeoffs and the ThroughlineThe capstone of the Orca case study: the engineering judgment behind every choice, and the one idea that ties them all together.8 sections