Hyperbolic and Geometric Deep Learning

I study neural architectures that exploit non-Euclidean geometry, with a focus on hyperbolic representations for hierarchical and relational data. My work explores both theoretical properties and practical performance of geometry-aware networks.

Theoretical Properties of Transformers

I investigate theoretical aspects of Transformer architectures, including expressivity, approximation properties, and geometric variants. This includes developing hyperbolic extensions and analyzing their consistency and representational power.

Operator-Theoretic Perspectives on Learning

My research applies tools from operator theory and functional analysis to understand neural networks as operators between function spaces, aiming to bridge classical mathematical theory with modern deep learning models.

Deep Reinforcement Learning for Autonomous Trading

I explore reinforcement learning frameworks for autonomous trading systems, focusing on stability, risk-aware decision-making, and learning in non-stationary financial environments.