Manas Sahni

I'm a Systems Software Engineer at NVIDIA. I work on GPU kernels for AI, compilers to make them easier to write, and libraries to make them easier to use. I currently work on CUTLASS, where I'm building the CUTLASS Operators project.

I completed my MS at Georgia Tech, where I was advised by Alexey Tumanov in the Systems for AI Lab, working on hardware-aware neural architecture search, real-time machine learning inference, and AutoML. I also interned with NVIDIA TensorRT, optimizing end-to-end systems for inference on GPUs.

Earlier, I was an ML Engineer at Samsung India R&D, where I helped bring flagship vision applications to low-power devices, through the Samsung Neural SDK, the first smartphone NPUs, quantization, and lightweight CNNs.
I completed my Bachelor's degree in Math & Computing at Delhi Technological University.