Can you tell us more about your role and how you apply data science to your projects?

I am a part of the Human Aligned Video AI (HAVA) Lab and work at the intersection of AI and dentistry, focusing on how machine learning can transform dental education and diagnostic training. Traditional methods for teaching caries detection and pathology identification often lack the standardization and personalized feedback that learners need to develop accurate diagnostic skills at scale. I apply data science through a patient-centric lens, building AI systems that prioritize real-world clinical impact and measurably improve patient outcomes.

Is there a project from this past year that you are most proud of?

I'm most proud of a collaborative project with another DSC PhD student where we're developing more robust medical image segmentation models that can handle real-world clinical variability. In practice, clinicians often deploy pretrained models without retraining due to resource constraints, which means these models must handle distribution shifts like different CT scanners, varying noise levels, or changes in image contrast. By rethinking the underlying geometry of standard neural network architectures, specifically exploring hyperbolic geometry, we've created models that are significantly more robust to noise, adversarial attacks, and brightness variations, ultimately making AI tools more reliable and generalizable in actual clinical settings.

What do you like most about being a DSC member?

What I value most about DSC is the remarkable diversity of research applications. While we're all working in data science, every conversation reveals someone tackling a completely different domain, which constantly broadens my perspective. The community has been incredibly supportive, organizing engaging events, talks, and training sessions that are both informative and genuinely enjoyable. I also had the opportunity to organise the DSC Away Day, along with other PhDs from the HAVA Lab this year and it was a great experience.

What is your favourite data science method?

I'm particularly drawn to geometric deep learning methods, especially those incorporating hyperbolic geometry into neural network architectures. If I had to choose one method for this month, it would be the “focal loss”. It addresses class imbalance with elegant mathematical formulation that's just a simple modification of cross-entropy.

Are you camp Python/R/or something else?

I will always remain loyal to Python (Pytorch! NOT Tensorflow).