AI/ML Scientist — Medical Imaging

Black Swans Exist; København; On-site; DKK 850,000 - 1,100,000; Full time

Job summary

Black Swans Exist in Copenhagen is seeking its first AI/ML scientist to help shape diagnostic tools based on PET imaging data. You will collaborate with clinicians at partner hospitals to define problems, build prototypes and test ideas in real clinical settings.

The role offers substantial ownership, travel within Denmark and nearby regions, and close collaboration with eight software developers reporting to the Head of Software.

Qualifications

• AI/ML development experience with tangible projects
• Software engineering and algorithm background with PyTorch, TensorFlow, JAX or Keras, disciplined testing and version control
• Experience with data collection, selection, integration and quality control, and preparing datasets for training
• Comfort starting from a rough idea rather than a formal spec
• Willingness to engage clinicians; not working from a desk
• 3–8 years of relevant experience; PhD or PhD + postdoc counts if applied and demonstrable
• Medical imaging knowledge and 3D/4D data experience is a plus

Responsibilities

• Invent, prototype and mature algorithms for extracting clinically meaningful information from PET data
• Translate clinical and scientific questions into testable algorithmic solutions
• Build pipelines around the work — acquisition, quality control, processing, visualization, presentation of findings
• Go to hospitals; observe workflows; ask clinicians what they need and validate assumptions against clinical reality
• Evaluate emerging methods and tools and shape investment directions
• Support AI/ML aspects of medical device compliance towards CE-marking and FDA

Job description

AI/ML Scientist — Medical Imaging

Copenhagen · Permanent · Presented by Black Swans Exist

The short version

A venture-backed Danish medical imaging company is hiring its first dedicated AI/ML scientist. The company has a CE-marked product already in clinical use and is now building the next generation of diagnostic tools on top of the imaging data that product generates.

This is not a role where you inherit a dataset and a backlog. The data largely does not exist yet. Part of this job is going out to hospitals, working alongside clinicians, and finding out whether the data they hold actually contains the answer to the question they are asking. The other part is turning that into methods and working prototypes.

If you want a clean pipeline and a defined problem, this is the wrong role. If you want to define the problem, it is close to ideal.

The company

Venture-backed, Danish, around 40 people, with offices in Copenhagen. The technology is PET-based functional imaging that produces quantitative measurements rather than pictures alone — the ability to put a number on physiological function, not just describe what an image looks like.

The first product is CE-marked and in clinical use, with a pivotal trial running and a regulatory submission planned in the United States. Quantification is the core of the company's value: it is what competitors do not offer, and it is the thing this role is hired to extend.

The company is named at first conversation. We have not put it in writing here.

The mission

The imaging data is the future product line. Today the company measures blood flow in the heart. The question is what else the same data can answer - in cardiology, in neurology, and beyond - and what methods are needed to get there.

You are hired to answer that. You will be the company's dedicated AI/ML voice, joining a software team of eight and reporting to the Head of Software. You will work across R&D, clinical development, quality and regulatory.

Concretely, the first twelve months look like this:

• Building working relationships with clinicians and technicians at partner hospitals, in Denmark and nearby
• Getting into clinics and understanding how the scans are actually acquired and read
• Identifying which data holds real signal — and being honest when it does not
• Running studies and projects alongside hospital doctors
• Building prototypes that show whether a method works

Success at six months is measured in relationships and prototypes, not in shipped features.

What you will actually do

• Invent, prototype and mature algorithms for extracting clinically meaningful information from PET data
• Translate clinical and scientific questions into testable algorithmic solutions
• Build the pipelines around the work — acquisition, quality control, processing, visualization, presentation of findings
• Go to hospitals. Observe workflows. Ask clinicians what they need and validate that your assumptions survive contact with clinical reality
• Evaluate emerging methods and tools, and shape which directions the company invests in
• Support the AI/ML dimension of medical device compliance as methods mature toward product — CE-marking and FDA

Travel is roughly two to three days per month, primarily within Denmark and southern Sweden.

Who we are looking for

Essential

• Demonstrable AI/ML development experience — applied, and with something you can point to and say "I built that"
• A software engineering and algorithm background, hands-on with PyTorch, TensorFlow, JAX or Keras, and disciplined about testing and version control
• Experience with data collection, selection, integration and quality control, and preparing datasets for training
• Comfort starting from a rough idea rather than a specification
• Someone who wants to be in the room with clinicians. This role does not work from a desk — you need to be willing to initiate conversations with doctors and technicians, ask uncomfortable questions, and keep asking
• Roughly three to eight years of relevant experience. A PhD, or a PhD plus postdoc, counts — provided the work was applied and you have built things

Strongly preferred

• Medical imaging: DICOM, and work on 3D or 4D volumetric data — segmentation, registration, classification or quantification. PET, MR or X-ray all count
• PET-specific knowledge — reconstruction, attenuation and scatter correction, tracer kinetics, quantification. Genuinely valuable, genuinely not required on day one. The company expects to teach this
• Medical device software or SaMD experience — IEC 62304, quality systems, verification and validation, CE-marking or FDA
• A background in healthcare, medtech, diagnostics or another regulated field

Education

Computer science, biomedical engineering, medical physics or similar is typical. Proven capability substitutes for credentials.

Relocation

Candidates relocating to Denmark will be considered.

What this is, honestly

The genuine appeal.

This is the company's first and only AI/ML hire. Nobody has done this before you, there is no established way of working to inherit, and the direction the company takes on AI is the direction you argue for. It is an unusual amount of ownership for someone at this level, and it is the main reason to take the role.

What you should weigh.

You will be the only AI/ML person among eight developers. There is no in-house peer to review your methods and no senior imaging-AI person to learn from — the clinical and scientific expertise sits with the company's hospital collaborators and advisors rather than inside the building. Some people find that liberating. Others find it isolating, and it is worth being honest with yourself about which you are before you apply.

The other thing to weigh is the data. It is early. Expect to spend real time on access, agreements, site visits and data quality before you spend it on model architecture.