Ravi Gupta is a Senior AI Models Researcher at AMD, where he leads open research on serving large AI models across GPU clusters. As principal investigator on AMD’s Disaggregated Inference series, he has scaled frontier models such as GLM-5.1 and the 2.8-trillion-parameter Kimi-K3 across multi-node AMD Instinct deployments. He also developed a method for running 4-bit (MXFP4) models at FP8 accuracy and speed on GPUs without native 4-bit support. Earlier in his career, he worked with Argonne National Laboratory on Aurora, its exascale supercomputer, leading research on how deep learning scales across thousands of GPUs. His research began at Purdue University and Lawrence Livermore National Laboratory in 2014 with STATuner, a machine-learning method for GPU kernel tuning. He has published 18 peer-reviewed papers, serves on program committees including ICPP, and is an IEEE Senior Member.