**Summary: **

• AI Foundation models are increasingly being applied in the medical field, specifically radiology, to maximize state-of-the-art results while also bringing attention to biases. RayDINO, a large visual encoder, has been developed and trained with over 873k chest X-rays. This model was evaluated across nine radiology tasks which include classification, dense segmentation, and text generation. **Findings: **
• The self-supervised approach helps in reducing biases associated with patient age, sex, and population. **Significance: **
• The application of self-supervision in medical AI aids in improving clinical workflows and the holistic interpretation of X-rays, proving critical for patient-centered care. **Potential Applications: ** Further research can explore self-supervised models in other medically relevant imaging techniques.