���������¼�

Hello and Welcome

Hi there — I’m Nic! I’m generally interested in the cross-section of machine learning, mathematics, and statistics — especially when these come together to help us reason about complex systems like healthcare, language, or the brain. I enjoy asking questions that sit somewhere between theory and application: How should we represent uncertainty? Can machines generate meaningful data? What makes a model interpretable, or even trustworthy?

Much of my recent work explores how synthetic data can support these kinds of questions — both technically and ethically. To me, synthetic data isn’t just about creating “fake” records — it’s about designing usable, testable, and often elegant approximations of real-world phenomena. That opens up exciting possibilities across three main areas:

• Synthesis – Using synthetic data to make healthcare datasets more publicly accessible, while protecting privacy and confidentiality.
• Augmentation – Generating synthetic examples to address inherent biases or imbalances in downstream models.
• Imputation – Understanding the role of missing data handling and how different practices influence the trajectory of research and results.

Each of these areas has both practical and philosophical implications — and that’s what keeps me coming back.

On Memory, Latent Spaces, and Representation

Beyond data generation, I’m also fascinated by questions of memory and latent space representations in machine learning. I often find myself drawn to how models store, compress, and recall information across time and tasks.

What does it really mean for a model to “remember”? Can we think of latent spaces as blueprints of knowledge — or maps of the world? And how do we design models that not only perform well, but also help us understand the systems they’re trained on?

This is the space where theory meets intuition — and where I find some of the most inspiring challenges in modern AI.