Sam Estep joined NVIDIA as a research intern in summer 2025, working on the design of language and compilers for heterogeneous graphics programming (seamlessly mixing CPU and GPU code). Sam also has extensive experience in building differentiable programming systems, with projects like Floretta (a source-to-source auto-diff tool for WebAssembly) and Rose (a composable JS-based auto-diff tool) that form the backbone of the Penrose layout engine. Sam’s latest work is GradBench, which benchmarks the performance of >23 auto-diff frameworks on intensive workloads.

He is currently a Ph.D. student at Carnegie Mellon University’s S3D and is advised by Joshua Sunshine.