namusanga

Location: Kampala, Uganda

Remote: Yes, I've worked remotely with UK and US companies

Willing to relocate: Yes, if the role needs it

Technologies: CUDA, C++, PTX/SASS level tuning, Nsight Compute, Nsight Systems, cuRAND, cuSPARSE, sparse linear algebra, Vulkan, GLSL, Slang, NVIDIA DeepStream, Unreal Engine, HPC, physics simulation, Python

Résumé/CV: https://namusanga.com/resume/nicholas_namusanga_cuda_kernel_...

Email: <email>

I make GPU workloads faster. I profile the hot path, find where memory traffic, occupancy or launch setup is costing throughput, and rework the kernels until the numbers move.

Recent work:

Optimizing quantum Monte Carlo sampling in CUDA. The sampler runs 26x faster than the standard cuRAND implementation. Article: https://namusanga.com/blog/quantum-monte-carlo-cuda Code: https://github.com/nnamu-cl/cuda-orbital-sampler

psiEngine, a physics and simulation engine on the GPU: CUDA compute feeding a Vulkan renderer, with custom sparse linear algebra kernels. https://github.com/nnamu-cl/psiEngine

Real time AI video for the Jugo platform. I optimized the Unreal renderer and video pipeline on an AWS cluster running RTX PRO 6000 Blackwell GPUs. I can only share the public showcase: https://www.jugo.io/videos

Behind that, 10 years of low-level C++ across HPC, graphics and simulation.

I'm open to contract or full time. With a lot of my previous companies the pattern has been to start on contract to deliver a milestone, then move into a steady role. I'm happy to have a discussion on specifics of any of the work.

  Location: Kampala, Uganda

  Remote: Yes (have worked remotely with US and UK contractors and open to on-site)

  Willing to relocate: Yes

  Technologies: CUDA, PTX/SASS level kernel tuning, C++, Nsight profiling, TensorRT, HPC, physics simulation, sparse solvers

  Résumé/CV: https://namusanga.com

  Email: [email protected]

GPU kernel engineer. With focus and experience in CUDA, HIP/ROCm and kernel level optimization. Excel at profiling and optimizing GPU compute to target peak performance and exploit platform and hardware specific advantages.

Recent work: CUDA Quantum Sampling kernel running ~26x over cuRAND (Article: https://namusanga.com/blog/quantum-monte-carlo-cuda), GPU physics solver and renderer (https://github.com/nnamu-cl/psiEngine), real time AI video processing on RTX PRO 6000 Blackwell GPUs for a production platform (showcase: https://www.jugo.io/videos)

Open to contract or full time. Happy to start with bounded and integrate gradually.