Scott Gray is an independent AI researcher exploring neuro-inspired approaches. Previously, he worked at OpenAI on GPU kernels, low-level compute optimization, sparse transformers, and large-scale language-model training.
Scott Gray is an independent AI researcher exploring neuro-inspired approaches. Previously, he worked at OpenAI on GPU kernels, low-level compute optimization, sparse transformers, and large-scale language-model training.
Career
Member of Technical Staff / Research Engineer at OpenAI from 2016 to PresentDeveloped fast GPU kernels and low-level software optimizations for AI compute. Contributed to sparse transformers, large-scale language-model training, GPT-4, and GPT-4o-related work. Worked on GPU kernels for block-sparse weights and neural-network performance.