Anish's vision is to rethink data movement across the Large Language Model stack to improve LLM training and inference efficiency. He aims to overcome the GPU memory wall by adapting LLM architectures and system software to maximize data reuse. His longer-term goal is to architect disaggregated memory systems that scale the GPU memory capacity. His ongoing research includes accelerating emerging LLMs in latency-critical serving scenarios and designing intelligent memory pools to support the deployment of larger LLMs. Anish has previously worked on designing CXL-centric server memory systems and DRAM subsystems resilient to data-disturbance errors.
Anish Saxena is a fourth-year PhD student working with Prof. Moinuddin Qureshi at Georgia Tech. He obtained his Bachelor of Technology in Mechanical Engineering from IIT Kanpur in 2021, advised by Prof. Biswabandan Panda. His professional experience includes internships at Intel Labs, Micron, AMD Research, Nvidia Research, and collaborative projects with Qualcomm and NXP Semiconductors. Previously, he was awarded the Aditya Birla Group Scholarship (Class of 2017) and the Kishore Vaigyanik Protsahan Yojana (KVPY) Fellowship in 2017. Outside work, Anish is a movie buff, dancing enthusiast, tea and coffee aficionado, and enjoys performing arts and hiking.