Sepehr Abdous

I design practical systems for high-performance and reliable datacenter networks. My research spans congestion detection and mitigation, packet routing, and efficient communication within and across datacenters. I also work on designing efficient and scalable networking solutions for distributed AI systems.

Education

2019 — 2026

Johns Hopkins University

Ph.D. in Computer Science

Baltimore, MD

Advisor Dr. Soudeh Ghorbani

Thesis Taming the Beast: In-network Congestion Management for Large-scale Datacenters

2019 — 2023

Johns Hopkins University

M.Eng. in Computer Science

Baltimore, MD

Advisor Dr. Soudeh Ghorbani

GPA 3.8 / 4

2014 — 2019

Sharif University of Technology

B.Sc. in Computer Engineering

Tehran, Iran

GPA 17.06 / 20

Research

My research focuses on building more efficient networking infrastructures. I seek practical solutions to pressing challenges that limit the performance and scalability of modern networks, particularly those supporting demanding applications such as large-scale distributed AI training. A few of my publications:

All publications →

Closing the Bandwidth Gap in AI Datacenters with Scalable Multicast

Sepehr Abdous*, Jinqi Lu*, Sana Mahmood, Jiacheng Wan, Alex Minetree Dill, Erfan Sharafzadeh, Ying Zhang, Soudeh Ghorbani · *Equal contribution

NSDI 2027

Uno: A One-Stop Solution for Inter- and Intra-Datacenter Congestion Control and Reliable Connectivity

Tommaso Bonato*, Sepehr Abdous*, Abdul Kabbani, Ahmad Ghalayini, Nadeen Gebara, Terry Lam, Anup Agarwal, Tiancheng Chen, Zhuolong Yu, Konstantin Taranov, Mahmoud Elhaddad, Daniele De Sensi, Soudeh Ghorbani, Torsten Hoefler · *Equal contribution

SC 2025

Practical Packet Deflection in Datacenters

Sepehr Abdous, Erfan Sharafzadeh, Soudeh Ghorbani

CoNEXT 2023