Research Scholar at IIT Bombay, Mumbai.

I am a PhD student in the Computer Science and Engineering Department at Indian Institute of Technology Bombay, advised by Prof. Mythili Vutukuru. I am a member of the Synerg Lab, where my research focuses on the intersection of systems, networking, and systems for machine learning.

Recently, I have been focusing on modern LLM serving systems such as vLLM and SGLang, aiming to understand and mitigate performance bottlenecks in real-world inference workloads. My ongoing research explores new design opportunities in scheduling, memory offloading, and network-level optimizations to enable efficient large-scale AI inference. While exploring inference systems, we developed a prefix-aware scheduler that uses reinforcement learning (RL) to learn the optimal trade-off between batch size and prefix homogeneity Feather. More recently, I am exploring Mixture-of-Experts serving systems, focusing on expert-aware request batching and scheduling, while profiling performance on the vLLM platform.

During my PhD seminar, I explored the programmability of the Linux network stack, which plays a pivotal role in granting developers the ability to modify, extend, and optimize networking behaviors within the Linux operating system. I reviewed several research works showcasing the possibility of programmability at different levels of Linux Operating System Report.

While exploring Linux network stack, I examined software kernel-bypass frameworks such as AF_XDP and hardware optimizations like DDIO (Data Direct I/O). In my work Toasty, we addressed the well-known leaky DMA issue in DDIO limitation by designing a cache-aware buffer management strategy that adapts to traffic dynamics, improving both performance and robustness, available in the repository.

Academics

News

  • Mar 2026 - I attended and presented my work Toasty at ASPLOS ‘2026 held in Pittsburgh, USA.
  • Feb 2026 - I volunteered in a hands-on session on docker internals, kubernetes and eBPF at IIIT Delhi. Linkedin
  • Feb 2026 - Our paper “Toasty: Speeding Up Network I/O with Cache-Warm Buffers” got accepted at ASPLOS ‘2026. Linkedin, Linkedin
  • Jan 2026 - I attended Google India Networking Research Summit 2026 held at Google Bengaluru office.
  • Sept 2025 - I participated in Workshop on Systems for LLMs organised by IBM, IIT Bombay. Link
  • Mar 2025 - I presented poster on “Improving Network I/O Performance Using Cache-Warm Buffers”, in RISC 2025 at IIT Bombay. Link
  • Jan 2025 - I received Asha Navani PhD Fellowship. Linkedin