wlruys

Definitely the market incentives for "good code" have never been worse, but I'm wouldn't be so sure the cost of migrating decent pieces of generated code to good code is worse than writing good code from whole cloth.

Just to add to this, I think John Levine's Linkers and Loaders is also a great reference.

Location: Austin, TX (targeting Northeastern US: NJ, NY, MA, NH)

Remote: Open to remote (prefer hybrid)

Willing to relocate: Yes

Technologies: Python, C++17, CUDA/HIP, PyTorch (TorchRL, Geometric), MPI, OpenMP, Kokkos, BLAS/LAPACK, SQL, Ray, Slurm, Docker/Singularity, distributed training, profiling (Tracy, perf, NVTX/Nvidia Nsight), Nanobind, Cython

Resume/CV: https://linkedin.com/in/wlruys (PDF available on request)

Email: [email protected]

PhD in high-performance computing, focused on algorithms and orchestration on heterogeneous clusters. I've worked on developing runtime systems, GPU kernel & simulation optimization, image registration, and nearest neighbor algorithms. My current research applies reinforcement learning (RL) to dynamic scheduling decisions.

Seeking ML infrastructure, distributed systems, or software performance roles in industry. Also open to RSE positions at research-focused organizations. Primarily targeting the northeastern US but flexible on location for the right opportunity, including other US hubs or Europe.