Context and Issues
In the field of medical imaging, manual annotation of 3D images represents a major challenge, requiring considerable time and specific expertise from radiologists, which limits and slows down the development of advanced clinical AI models. The emergence of foundational models for medical imaging in computer vision, such as VISTA3D, opens up new perspectives for supporting annotation tasks.

Internship Objectives
Evaluate the capabilities of a foundational model for medical imaging and develop an interface in order to evaluate assisted annotation process (speed, quality). 

The project will be structured in the following way:

• State of the art of medical foundational models
• Simple and semi-supervised inferences with chosen model
• Development of the visualization/annotation interface
• Evaluation of annotations produced for training internal segmentation models

Developed Skills

• Deep learning: Self-supervised learning, semi-supervised learning, segmentation models
• Libraries: Torch, Monai, Trames, Dash, Pyvista
• Cloud: AWS
• Deep learning in a medical context
• Medical imaging

• Strong foundations in Python and software development (must have)
• Knowledge in machine learning and Deep Learning (must have)
• Interest in medical AI applications, particularly computer vision
• Rigorous and proactive (must have)
• Experience with “interface” libraries (Trames, PyVista, Streamlit or others)
• Initial experience (internship / project) in a start-up environment (nice to have)

Level: Master’s Degree (final year) / Engineering school
Duration: 4-6 months