SpatialX: Multi-Modal Annotation System
As CTO of SpatialX AI (2024–2025), I architected a multimodal AI platform for digital pathology end-to-end, under HIPAA and ISO 15189, hands-on across frontend and backend.
I led the SpatialX patent from concept to delivery: an AI-powered Multi-Modal Annotation System, plus 5 research papers. The pipeline serves whole-slide images at up to 1M+ image-tiles per minute.
Tech: multimodal AI, whole-slide imaging (WSI), tile-streaming pipelines, AWS, Pulumi/IaC, CircleCI.
UX — An Annotation Tool That Feels Like ProCreate on iPad
Pathologists do not think in form fields and dropdowns — they think with their hands. So I built the annotation surface to feel like ProCreate on iPad: direct, low-latency brush strokes on gigapixel slides, pressure-aware marking, and pan/zoom that never fights the user.
The goal was zero friction between an expert's intent and the label that trains the model. Annotating a region should feel like sketching, not data entry. The recording below shows the intuitive annotation tool in action on a whole-slide image.