Arjun Devraj

PhD Candidate · Cornell University

adevraj [at] cs [dot] cornell [dot] edu

I'm a third-year PhD candidate in Computer Science at Cornell University, where I'm advised by Rachee Singh. My research interests are broadly in systems and networking, with a focus on distributed machine learning and wide-area networking. I'm particularly interested in using new optical and photonic technologies to overcome communication bottlenecks in distributed machine learning systems. My work is supported by the NSF Graduate Research Fellowship.

Before graduate school, I spent two years as a software engineer at Meta, where I worked on privacy infrastructure for the ads recommender system. Previously, I received my bachelor's degree in Computer Science, summa cum laude, from Princeton University, where I worked with Jennifer Rexford on network privacy and with Tom Griffiths and Qiong Zhang on computational cognitive science.

Arjun Devraj

Recent News

Publications

Autonomy Without Anarchy in Decentralized WAN Traffic Engineering
[pdf]
Arjun Devraj, Umesh Krishnaswamy, Ying Zhang, Karuna Grewal, Justin Hsu, Éva Tardos, Rachee Singh
USENIX NSDI, 2027
Mo Collectives Mo Problems
[pdf]
Abhishek Vijaya Kumar, Arjun Devraj, Rachee Singh
USENIX NSDI, 2027
λλ: A Programming Language for Silicon Photonics
Vaibhav Mehta, Arjun Devraj, Bill Owens, Justin Hsu, Rachee Singh
ACM SIGCOMM, 2026
Reconfigurable Torus Fabrics for Multi-tenant ML
[pdf]
Abhishek Vijaya Kumar, Eric Ding, Arjun Devraj, Darius Bunandar, Rachee Singh
ACM ASPLOS, 2026
In-Network Analog AllReduce for ML with Programmable Integrated Photonics
[pdf]
Arjun Devraj, Bill Owens, Daniel Pérez-López, Rachee Singh
Optical Fiber Communication Conference (OFC), 2026 Oral Top Scored
HEDGE: Traffic Engineering with Probabilistic Link Capacities
[pdf]
Arjun Devraj, Bill Owens, Umesh Krishnaswamy, Ying Zhang, Rachee Singh
USENIX NSDI, 2026
The Reality of Chasing Shannon's Limit in Optical Wide-Area Networks
[pdf]
Arjun Devraj, Bill Owens, Rachee Singh
ACM Internet Measurement Conference (IMC), 2025
Chip-to-chip photonic connectivity in multi-accelerator servers for ML
[pdf]
Abhishek Vijaya Kumar, Arjun Devraj, Darius Bunandar, Rachee Singh
Optical Fiber Communications Conference (OFC), 2025 Oral
A case for server-scale photonic connectivity
[pdf]
Abhishek Vijaya Kumar, Arjun Devraj, Darius Bunandar, Rachee Singh
ACM HotNets, 2024
Reconciling categorization and memory via environmental statistics
[pdf]
Arjun Devraj, Thomas L. Griffiths, Qiong Zhang
Psychonomic Bulletin & Review, 2024
REDACT: Refraction Networking from the Data Center
[pdf]
Arjun Devraj, Liang Wang, Jennifer Rexford
ACM SIGCOMM Computer Communication Review (CCR), 2021
The Dynamics of Exemplar and Prototype Representations Depend on Environmental Statistics
[pdf]
Arjun Devraj, Qiong Zhang, Thomas L. Griffiths
Proceedings of the 43rd Annual Conference of the Cognitive Science Society, 2021
Efficient AllReduce with Stragglers
[pdf]
Arjun Devraj, Eric Ding, Abhishek Vijaya Kumar, Robert Kleinberg, Rachee Singh
Preprint

Talks and Presentations

Efficient AllReduce with Stragglers
  • Google Networking Research Summit (October 2025)
  • ACE Center for Evolvable Computing (August 2025)
Stable and Fault-Tolerant Decentralized WAN Traffic Engineering
  • Meta Workshop on WAN Control (October 2025)
HEDGE: Traffic Engineering with Probabilistic Link Capacities
  • Workshop on Reconfigurable Networks (June 2025)
The Reality of Chasing Shannon's Limit in Optical WANs
  • Cornell Systems Research Seminar (March 2024)