About Me
PhD Student in Machine Learning
I explore the boundaries of graph machine learning and neural reasoning.
I investigate how to build better graph learning architectures and teach neural networks to reason properly. More broadly, I am interested in how we can achieve generalization, particularly in extrapolation settings using reasoning and computation depth.
I completed my studies at ETH Zurich and joined the DISCO Lab in 2022 under the supervision of Professor Roger Wattenhofer.
Research Interests
-
What are the boundaries of graph machine learning?
I investigate the limits of graph learning architectures, what makes certain tasks fundamentally hard, and how architectural choices shape what and how well models can learn. -
How can we teach models to generalize?
Classical algorithms generalize well but are hard to design or even learn, while machine learning models learn easily but often fail to generalize. I aim to explore the intersection through neural reasoning: how we can incentivize it, and leverage more computation to generalize to much larger and more complex instances beyond the training distribution. -
How can we capture and quantify long-range interactions?
I am interested in methods that go beyond local-only information to mimic reasoning or real-world systems that require long-range interactions. I try to quantify progress towards this by designing appropriate benchmarks and proposing architectures.
CV
For a detailed overview of my education, experience, and publications, see my full CV.
Download CV (PDF)
