Clinical Imaging Platform Engineer

We build clinical AI that reads alongside radiologists. Our abdomen-pelvis CT triage device is FDA-cleared, and it's the first commercial system to simultaneously triage seven urgent conditions on abdomen-pelvis CT in the U.S. We're backed by Khosla Ventures.

About the role

This role owns two critical factors that determine the quality of our models: the accuracy of the labels we use for training, and the usability and trustworthiness of the model's output for radiologists. Radiologist time is our most valuable resource, and inefficiencies in the interface directly impact annotation throughput, model quality, and ultimately patient outcomes. This position blends product engineering with infrastructure, as the interface radiologists use directly influences model performance.

The core challenge is ensuring label quality. Our annotation pipeline is designed to measure itself: cases are assigned without race conditions, annotators follow defined phases, some batches include known ground truth, others are intentionally reviewed by multiple readers, and we track agreement using per-lesion Dice scores—even when readers work from different reconstructions. Getting this measurement right is the majority of the work.

Currently, one engineer manages this entire surface while juggling other responsibilities. This hire will close that gap.

Responsibilities

• Own the viewer, built on Cornerstone3D and VTK.js, including:
  • Unified volume-rendering with fallback to stack rendering
  • Multiplanar reformats generated on demand in any plane, clearly labeled
  • Progressive loading for immediate low-resolution previews from the first 10% of HTJ2K codestreams
  • Radiological correctness (left/right orientation, gantry tilt rulers, MONOCHROME1 inversion, signed-pixel codec mismatches)
• Develop and maintain the annotation system, including:
  • 3D mask storage and AI-assisted segmentation tools (click-to-segment across planes, region growing, FWHM thresholding, multi-seed refinement)
  • Evaluation pipeline: case-pool ledgers, per-annotator phase state machines, ground-truth and peer-overlap batches, agreement scorecards
  • Cross-series resampling through NIfTI affines
  • Transparent handling of unscorable cases (no default zeros)
• Design and refine annotation schemas for per-lesion categories, gated by view, phase, and slice, with conditionally required fields—collaborating with radiologists on clinical decisions.
• Ensure model output is interpretable by radiologists:
  • Versioned sparse-RLE mask contracts keyed by SOP Instance UID
  • Geometry mapping from model grids to world-space contours (image position, direction cosines, pixel spacing)
• Maintain the platform infrastructure:
  • FastAPI on AWS with per-study authorization, cohort isolation, and HIPAA-compliant access auditing
  • Debugging DICOM geometry issues (coordinate systems, orientation, affines)

Requirements

• Experience building software used intensively (hours daily) and improving it through user observation.
• Instinct to measure quality with metrics, especially for annotation accuracy.
• Comfort with coordinate systems to debug precision issues (e.g., 3mm mask offsets) rather than manual adjustments.
• Commitment to fail-safe defaults when handling patient data.
• Ability to collaborate with radiologists: challenge them on software, defer to them on medicine.

Skills

Helpful but not required:

• Strong React and TypeScript with performance optimization experience
• Backend API and async development
• Medical imaging (DICOM, Cornerstone3D, OHIF, PACS)
• Annotation tooling (either development or usage)
• AWS and Terraform
• Segmentation or computer vision
• Inter-rater agreement and measurement design
• Regulated software experience (ISO 13485, IEC 62304, HIPAA)
• Codec or streaming work (e.g., HTJ2K)

Stack

• Frontend: React, TypeScript, Vite, Cornerstone3D, VTK.js, Jotai, TanStack Query, Tailwind, Playwright
• Backend: Python, FastAPI, pytest
• Infrastructure: AWS (DynamoDB, S3 byte-range reads, Cognito, SQS, ECS, Lambda) with Terraform
• Data: DICOM, DICOMweb, HTJ2K, NIfTI, RLE

Pay

• US (Boston hybrid): $140,000–$180,000 base, plus ~10% discretionary bonus. Equity may be offered to top candidates.
• International (remote): Cash-weighted base set by local market rates. No equity. Contractor or employer-of-record employment depending on country.

Schedule

Asynchronous, written reviews to accommodate global time zones. No two-tier engineering team—equal repository access, data, and review authority for all engineers.