Amit Rajaraman

Amit Rajaraman is a graduate student in electrical engineering and computer science, conducting research in theoretical computer science and the analysis of Markov chains. His work addresses a central puzzle: why do Markov chain–based algorithms often succeed in practical inference and optimization tasks, even when they lack formal mixing guarantees? Amit focuses on slow-mixing Markov chains and has introduced the framework of “locally stationary distributions” to analyze their convergence and performance. As a MathWorks Fellow, he will develop deeper connections between Markov chains and message-passing algorithms by designing theoretical frameworks and experimental comparisons. MATLAB supports this work by enabling visualizations, simulations, and empirical investigations that guide theoretical insights. Amit’s research could significantly reshape our understanding of optimization and inference in high-dimensional settings, providing a more principled foundation for commonly used algorithms, such as stochastic gradient descent. His efforts could ultimately advance both the theoretical foundations and practical applications of AI, machine learning, and statistical physics.