4/17/25 | 4:15pm | E51-376
Ali Aouad
Assistant Professor of Operations Management
MIT
Abstract: Managing “market thickness” is critical for digital service platforms and resource allocation matching markets, such as ride-hailing, organ donation, and emergency response. However, optimization-based matching policies are often computationally intractable in dynamic market environments. Previous research focuses on static policies or simplified market dynamics. I will present new results and insights into the design of control algorithms for two-sided matching queues. We develop near-optimal, efficient algorithms for spatial and bounded-size networks, revealing that the matchmaker can decompose the network into “short” and “long” queues and apply fundamentally different types of policies for each. Our hybrid LP technique combines dynamic programming in a frugal manner with drift analysis. Additionally, we achieve improved approximations for general networks by adopting a simple “proposal-based” market design, in which each agent proposes matches with fixed marginal probabilities. By modelling correlation among proposals, our algorithm can surpass the best-known performance guarantees. These findings offer effective new matching policies applicable to a range of dynamic, real-world markets. This is joint work with Alireza Amanihamedani (London Business School), Amin Saberi (Stanford), and Tristan Pollner (Stanford).
Bio: Ali Aouad is an Assistant Professor of Operations Management at MIT Sloan, specializing in algorithm design under uncertainty and data-driven decision processes in areas covering choice modeling, the design and operations of digital marketplaces, as well as public sector and social operations. His work was recognized by several awards, including the 2024 POMS Analytics Research Challenge Prize, the 2022 Second Prize of INFORMS Junior Faculty Interest Group, and a selection for funding in the European Research Council’s 2022 Starting Grant.