Weaves, Wires, and Morphisms
Formalising and implementing the algebra of deep learning, from mathematical expressions to executable models.
Researcher · MIT LIDS
Deep learning,
made legible
I study the mathematical structure of deep learning and how it connects to the hardware that runs it.
I’m a PhD student at MIT’s Laboratory for Information & Decision Systems, advised by Gioele Zardini. My work brings together category theory, neural network architectures, and GPU computing to make AI algorithms easier to understand and more efficient to run.
Formalising and implementing the algebra of deep learning, from mathematical expressions to executable models.
A diagrammatic approach to understanding data movement and the hardware behind efficient attention.
A visual language for communicating, implementing, and analysing deep learning architectures.
Investigate the DeepSeek-V4.1-Flash model, algebraically expressed
Explore the diagram