Computer Vision Engineer

Hugo Lindqvist

Full-time · Mid-level · Chicago

This role is no longer accepting applications

Similar open roles you can apply to now:

About the role

We make software that assists radiologists in reviewing chest X-rays and CT scans for pulmonary conditions — specifically early-stage pneumonia, pleural effusion, and pulmonary nodules. Our models are FDA-cleared and used in 12 hospital systems across the US. We are not an AI-replacing-the-radiologist company. We are an AI-helping-the-radiologist-process-a-growing-backlog company. That distinction matters to how we measure performance, how we communicate model uncertainty, and how we think about failure modes. The radiologist is always in the loop. Our job is to make their review faster and surface cases that warrant closer attention. We're looking for a mid-level computer vision engineer who understands that in medical imaging, a missed positive is a different kind of error than a false alarm — and that both must be measured, reported, and understood, not collapsed into a single accuracy number. You'll work within a team of six engineers and two clinical advisors. Clinical context is always part of the conversation here.

Responsibilities

Improve sensitivity and specificity of pulmonary nodule detection across active model versions
Build and maintain a clinically-grounded evaluation framework measuring sensitivity, specificity, PPV, and calibration
Collaborate with clinical advisors to translate radiologist workflow requirements into model specifications
Participate in model performance reviews with the regulatory affairs team and support submission documentation updates
Document all model changes in a format that supports regulatory audit trail requirements

Requirements

3–5 years of computer vision engineering with at least one year on medical or high-stakes imaging applications
PyTorch for model development, training, and fine-tuning — you write training loops and debug gradient issues independently
Deep understanding of object detection and segmentation for image analysis tasks
Experience with DICOM or other medical image formats: NIfTI, DICOM, PACS system concepts
Familiarity with model uncertainty estimation and confidence calibration
NumPy and medical imaging libraries (pydicom, SimpleITK, or equivalent)
Knowledge of FDA SaMD regulations or ISO 13485 quality management principles is a meaningful advantage

Benefits

Work where model outputs are clinically meaningful, measurable, and tracked against real patient workflows
Full remote with quarterly in-person clinical workshops
$100,000 – $125,000 base salary + equity
$2,000 annual conference and professional development budget
Access to de-identified clinical datasets from 12 hospital systems