AI/ML Engineering Intern, Medical Imaging
Learn to build the AI that reads radiology studies at DocOrbit. You'll take imaging models from prototype to production on real-world DICOM data, side by side with the founders and the radiologists who sign our reports.
About DocOrbit
DocOrbit is an AI-enhanced radiology second-opinion service. Patients and partner clinics send us MRI, CT, X-ray, ultrasound, mammography and PET-CT studies. Orbius, our AI radiology reporter, reads the full DICOM study and drafts a structured report that a board-certified radiologist can review and sign. Clinics also connect their PACS to us directly, so studies arrive exactly as hospitals produce them.
The role
As our AI/ML Engineering Intern, you'll work on imaging models from the DICOM file to the report. This is a hands-on engineering internship, not a research post: success means a model that runs reliably on real clinic data and a draft our radiologists trust. From your first weeks you'll own a real piece of our imaging pipeline and ship it to production, with the founders beside you.
What you'll do
- Build and harden the pipelines that turn raw DICOM studies into model-ready input: series selection, orientation, windowing and resampling across CT, MRI, X-ray, ultrasound and PET-CT
- Evaluate, adapt and deploy imaging models, from segmentation to finding detection, and feed their results into the reports Orbius drafts
- Combine classic imaging models with large vision-language models, so every draft rests on measurable findings
- Measure models against a gold set of radiologist-signed studies, catch regressions before they ship and turn radiologist feedback into improvements
- Run models in production on Kubernetes: package them, fit them to tight CPU and memory budgets, and monitor speed and failures
- Work with our PACS integrations (DICOMweb and DIMSE) so models meet studies where clinics already send them
- Handle patient data with care: de-identification, access control and audit logging are part of the job
What we're looking for
- A final-year, master's or PhD student, or an early-career graduate, in computer science, biomedical engineering, electrical engineering, physics or a related field
- Solid Python and hands-on PyTorch, from coursework, a thesis or your own projects
- Some hands-on work with medical images, ideally real DICOM series rather than only curated PNG datasets
- A practical builder: you'd rather ship a robust model than chase a leaderboard
- Comfortable with Git and the Linux command line
- Fluent written and spoken English
Nice to have
- Experience with medical imaging libraries such as pydicom, MONAI, SimpleITK or nibabel
- Segmentation work with nnU-Net, TotalSegmentator or similar
- Docker, APIs or cloud deployment, from a project or a previous internship
- LLMs or vision-language models applied to images or clinical text
What you'll learn
This internship is built around learning. You'll learn by doing, on real clinical data, with the founders as your mentors.
- How medical imaging AI works end to end, from a raw DICOM study to a report a radiologist signs
- How to evaluate models against radiologist-signed studies, not just benchmark scores, and catch regressions before they ship
- How to take a model from a notebook to production: packaging, Kubernetes, tight CPU and memory budgets, and monitoring
- How hospitals and clinics store and send images: PACS, DICOMweb and DIMSE
- How to handle patient data responsibly: de-identification, access control and rules such as GDPR, HIPAA and the EU MDR
- Real responsibility from day one, working directly with the founders and the radiologists who use what you build
Start your AI career in medical imaging
Email your CV to [email protected] with the subject “AI/ML Engineering Intern, Medical Imaging”, plus a link to your GitHub or a few lines on an imaging project you've worked on.