Research

My research lies at the intersection of distributed algorithms and systems implementation, with a focus on building resilient and scalable infrastructures. I am particularly interested in how emerging technologies such as data processing units in modern datacenters, reinforcement learning in multi-agent systems, and quantum resources for distributed coordination reshape the classical challenges of fault tolerance, consensus, and performance. Across these projects, my goal is to develop theoretically grounded algorithms that remain practical and impactful in real-world environments.

Current Projects

Frugal Distributed Algorithms at the Network Layer (FrugalDiNet)

Funding: French National Research Agency (ANR, 2024-2027)
Collaborators: Three French partner institutions

This project investigates how intelligent hardware, such as programmable switches and Data Processing Units (DPUs), can be leveraged for high performance services in large-scale datacenters. My focus is on identifying exploitable telemetry and designing distributed algorithms that enhance scalability and resilience at low cost. The work aims to establish a foundation for next-generation network devices attuned to the performance demands of modern infrastructure.

Scalable Collective AI: Learning and Decision-Making in Multi-Agent Systems

Funding: National Natural Science Foundation of China (NSFC, 2025-2028)
Collaborators: Dr. Mathieu Lauriere, NYU Shanghai

This project investigates the intersection of artificial intelligence, reinforcement learning, and mean-field game theory to understand how large populations of agents learn, adapt, and cooperate in complex environments. The research combines rigorous mathematical modeling with computational experimentation: mean-field games provide a framework for analyzing equilibrium behavior, while scalable algorithms are designed and tested to handle thousands of interacting agents. My contribution focuses on distributed simulation and algorithmic implementation, developing architectures that enable large-scale experiments, validate theoretical results, and extend them into tractable methods with applications in economics, finance, and network optimization.

QUANTum Improved COnsensus (QUANTICO)

Status: Exploratory project
Collaborators: Dr. Chandrashekar Radhakrishnan, NYU Shanghai

This collaboration examines the potential of entanglement to accelerate coordination in distributed quantum systems. We have introduced formal models that capture the interplay of quantum resources with communication delays and used them to design new approaches for leader election and consensus. Our findings suggest that quantum resources can significantly reduce complexity, but that delays remain fundamental, reframing classical impossibility results in the emerging context of hybrid quantum-classical computing.