Abhishek Vijaya Kumar

Abhishek Vijaya Kumar

PhD Candidate, Computer Science

abhishek [at] cs [dot] cornell [dot] edu
About me Publications Contact October, 2025

I am a Ph.D. candidate in the department of Computer Science at Cornell University where I am advised by Rachee Singh. My research spans both systems informed by current electrical interconnects and by emerging silicon photonics interconnects for ML accelerators. Specifically, my systems focus on addressing the memory capacity bottleneck and GPU resource underutilization by co-designing LLM inference engines with the scale-up network interconnecting GPUs (e.g., Nvlinks). Remaining projects in my thesis focus on addressing the interconnect bandwidth bottleneck in large direct-connect ML scale-up clusters (e.g., TPUs) using reconfigurable photonic fabrics. By reconfiguring the interconnect fabric, my systems (1) maximize the inter-accelerator bandwidth and (2) achieve near optimal collective communication performance. I am supported by the Bowers CIS-LinkedIn grant for the academic year 2024-2025.

Before joining Cornell, I had a wonderful time in the Systems group at Microsoft Research India. Here, I was advised by Dr. Muthian Sivathanu and Dr. Saikat Guha while working on mobile networks and storage stack for deep learning frameworks.Prior to that, I obtained my undergraduate degree in Information Science and Engineering from PES University, Bengaluru.

You can find my CV here: [PDF].

In my free time, I like to run 5Ks on this track, eat chaats with friends and play RPG games on PS5. My personal favorite is Dragon Ball Z sparking zero.

Publications

Aqua: Network-Accelerated Memory Offloading for LLMs in Scale-Up GPU Domains
Abhishek Vijaya Kumar, Gianni Antichi, Rachee Singh
ASPLOS, 2025

[PDF] [BibTex]

Chip-to-chip photonic connectivity in multi-accelerator servers for ML
Abhishek Vijaya Kumar, Arjun Devraj, Darius Bunandar, Rachee Singh
Optical Fiber Communication Conference (OFC), 2025

[PDF] [BibTex]

A case for server-scale photonic connectivity
Abhishek Vijaya Kumar, Arjun Devraj, Darius Bunandar, Rachee Singh
HotNets, 2024

[PDF] [BibTex]

Chisel: An optical slice of the wide-area network
Abhishek Vijaya Kumar, Bill Owens, Nikolaj Bjoner, Binbin Guan, Yawei Yin, Victor Bahl, Rachee Singh
NSDI, 2024

[PDF] [BibTex]

Quiver: An Informed Storage Cache for Deep Learning
Abhishek Vijaya Kumar, Muthian Sivathanu
FAST, 2020

[PDF][BibTex]

PCCL: Photonic Circuit-Switched Collective Communication for Distributed MLPREPRINT
Abhishek Vijaya Kumar, Arjun Devraj, Rachee Singh
arXiv preprint, 2025

[PDF] [arXiv]

Morphlux: Transforming Torus Fabrics for Efficient Multi-tenant MLPREPRINT
Abhishek Vijaya Kumar, Eric Ding, Arjun Devraj, Darius Bunandar, Rachee Singh
arXiv preprint, 2025

[PDF] [arXiv]

Contact

Please feel free to contact me or drop by my office if you are at Cornell.

Email: abhishek [at] cs [dot] cornell [dot] edu

Office: 456, New Bowers CIS building, Hoy Road, Ithaca, NY, 14853 .