Description
NVIDIA is seeking a Research Scientist to develop and optimize open-source inference frameworks, KV cache offloading systems, and distributed training performance technologies. The role involves building acceleration libraries and frameworks, analyzing computational bottlenecks, and contributing to community projects such as SGLang, vLLM, and FlexKV. Candidates need more than five years of technology-industry experience, a master’s degree or higher in a relevant field, strong programming and computer-systems knowledge, and interest in accelerated, parallel, and heterogeneous computing.
