Title: Stochastic Matching Networks: Theory and Applications to Online Marketplaces
Abstract: Owing to emerging applications, stochastic matching networks have received growing interest in the past decade. A defining feature is that both customers and servers arrive in the system and depart when matched. My research develops a theory for stochastic matching networks and considers its applications to online marketplaces, electric vehicle-based transportation systems, and payment channel networks. I will present two vignettes from this body of work in this talk.
First, motivated by applications to ride-hailing platforms, we study profit-maximizing pricing and matching policies in a bipartite matching network. We show that a two-price policy combined with a max-weight matching policy exhibits a cube root of n optimality rate when all the arrival rates scale by n. Second, I present the delay performance of a large class of pricing policies in heavy traffic in a single matching queue. Depending on how we approach the heavy-traffic limit, we observe a phase transition for the limiting stationary distribution of the number of customers and servers waiting in the system. I conclude my talk by providing an overview of my research interests and outlining future research directions.
Bio: Sushil Varma is a Ph.D. student in the ISyE department at Georgia Tech, advised by Prof. Siva Theja Maguluri. His research interests include stochastic processing and matching networks, game theory, and revenue management with applications in electric vehicles, online marketplaces, and load balancing. Sushil has received the INFORMS TSL Best Student Paper Award - Finalist 2023, the Stephen. S. Lavenberg Best Student Paper Award in IFIP Performance 2021, and the Alice and John Jarvis Best Student Paper Award in 2022.
