About

Hi, I’m Nicholas.

I build and optimize CUDA kernels and GPU pipelines for AI and HPC workloads. I work below the framework line: kernels written by hand, SASS read in Nsight Compute, memory traffic priced in bytes per element before the first line is written.

10+ years in low-level C++ on compute and memory bound workloads across AI compute, real-time video, simulation and numerical methods.

This site is where the write-ups live. Each project comes with its numbers, the profiler evidence behind them and what the next step would be. The GPU series covers the same ground one kernel at a time.

Open to remote GPU kernel and performance work, contract or full-time. Email hey@namusanga.com.