Geometric & topological deep learning · Computer vision · AI4Science
I am an Associate Professor at Imperial College London and a UKRI Future Leaders Fellow , where I lead the great CIRCLE research group and act as the elected AI speaker of Imperial's AI Network. Previously, I worked within the Geometric Computing Group at Stanford as part of StanfordAILab , and CAMP Group at TU Munich . I ask how intelligent machines should represent structure of the world by day, and indulge in music (drums and percussion) and philosophy by night.
Research · Questions and themes
What a machine can perceive, learn and explain depends on how the world is represented. My research programme has deliberately moved between 3D/4D visual perception, mathematics for machine learning theory, and scientific applications in physics, joined by shared underlying questions, I pose below:
Observations captured in the real world are messy and incomplete. Exploring their relation to geometry, symmetry and representations of shape and motion is key to create representations computers can understand.
The complexity of the real data cannot be captured only by simple grids, tables or pairwise graphs. Richer mathematical structures, such as manifolds and topological complexes, are essential to capture how systems truly interact.
The way we structure information dictates how an AI learns. We theoretically and empirically explore which models may reach reliable conclusions that don't break in new scenarios.
Scientific challenges in fields like physics, structural biology and chemistry require AI models that can truly adapt to any challenge. I recently started to build AI tools for molecules and proteins, modeling their form and interactions.
Research group
Publications
Book in development
With Mustafa Hajij, Theodore Papamarkou, Ghada Zamzmi, Karthikeyan Natesan Ramamurthy and Michael T. Schaub.
The book develops a common language for learning on simplicial, cellular and combinatorial structures beyond pairwise graph data.
Visit the book website →Teaching
Undergraduate and postgraduate teaching, doctoral supervision and shared reading spaces.
Teaching →Science beyond the lab
Seminars, keynotes, public lectures, summer schools and independent learning communities.
Outreach →Background
Outside research, I return to rhythm and philosophical inquiry as complementary ways of thinking about structure, expression and meaning. I play drums and percussion, sometimes joining Turkish indie-folk musician Zeyn’el Günbek, as well as the CVPR House Band.
Background and CV →
