CUDA Engineering Expert

What you'll do

The ad names a leading AI lab as the partner and GPU kernel optimisation as the focus. The work reads like a performance engineer's week:

• Profiling CUDA kernels and finding the bottleneck
• Proposing and implementing targeted optimisations for throughput on modern GPUs
• Refactoring C++ and CUDA code to be cleaner and portable across GPU architectures
• Writing GLSL and WebGPU shader logic and fitting it into existing pipelines
• Documenting each optimisation step and the measured gain

The combination of CUDA with GLSL and WebGPU is unusual. Most CUDA roles stop at compute; this one expects you to be comfortable in graphics shading and in the browser GPU API as well. How exactly the work trains a model (for example, whether you are producing reference optimisations for model-generated kernels) is not stated.

Who fits

• Demonstrated CUDA expertise with a record of tuning kernels
• Advanced C++ in high-performance settings
• Hands-on GLSL and WebGPU for graphics and compute shaders
• Fluency with GPU profilers such as Nsight or Visual Profiler
• The ability to reason about performance across hardware generations
• Clear technical reporting

Remote cross-disciplinary teamwork is a plus. No AI background is required.

What it pays

$60–100/hour, micro1's published band. For deep CUDA optimisation skill that is below typical industry rates, which suits people who want the work as a side engagement rather than a main income. No pay mechanism, hours or duration are stated. micro1 also lists a GPU Programming Software Engineer role at $60–95.

Worth knowing

Good:

• Real GPU optimisation work tied to a frontier AI lab
• 50 openings, marked high demand
• Rare chance to use CUDA, GLSL and WebGPU in one engagement

Less good:

• Pay under market for strong CUDA engineers
• The three-API mix narrows who fully qualifies
• No hours, duration or pay mechanism published
• Contractor engagement: no benefits, no notice period, your own tax to manage

About this listing

Posted by micro1 on its own jobs portal about a month before we read it on 24 September 2026. The band, the AI-lab reference and the skills are micro1's text; the lab is not named. Location is "Remote" with no country named. See micro1.