"Computers were made to automate tedious and error-prone tasks. Programming computers is a tedious and error-prone task. Why not just get a computer to do it?"
That question still frames how I think about programming language design, automatic programming, and performance software, from graduate research through my work at AMD.
I'm Bryant Nelson, Ph.D., a Principal Software Engineer, Technical Lead & Manager at AMD. As technical lead and engineering manager for GEMM Solution Selection in AMD Math Libraries on the ROCm stack (hipBLASLt / rocBLAS), I lead the team that chooses the right GEMM kernel for a given problem shape: the layer between user GEMM calls and the underlying kernel libraries. Our team owns and leads Origami, an analytical GEMM kernel-selection model that predicts near-optimal configurations from architectural parameters without runtime autotuning.
I earned my Ph.D. in Computer Science from Texas Tech University, focused on programming language design and automatic programming, and that mindset still shapes how I turn complex hardware into reliable, fast software. At IBM and now at AMD, I productize ideas from research into software that ships.
Side projects I maintain for fun and utility:
| Project | Description |
|---|---|
| board-game-chooser | Program to randomly choose a board game from my collection. |
| internet-troubleshooter | Python script for troubleshooting internet issues. |