Research

My research develops the theoretical and algorithmic foundations of optimization and reinforcement learning (RL), aiming to understand how model structures, stochasticity, and their interaction shape the behavior of modern learning algorithms. Drawing on optimization and statistics foundations, I study questions in policy optimization, statistical inference, and distributed training–developing methods that remain principled even in complex and uncertain environments. Since AI systems are increasingly deployed in critical environments, such principled approaches are essential for creating trustworthy, efficient, and scalable learning algorithms.

Key Research Directions

🔬 Non-convex Theory

Hidden Convexity Global Optimization Minima Selection

Mathematical frameworks for complex optimization landscapes.

⚡ Data Efficiency and Robustness

Heavy-Tailed Noise Policy Gradients Sample Efficiency

Reliable algorithms for challenging statistical conditions.

🚀 Scalable Systems

Distributed Training Communication Error Feedback

Communication-efficient distributed training at scale.