AI Generates Kernel Code Faster Than Human Engineers
CUDA engineers, known for their expertise in optimizing Nvidia chips, are increasingly supervising AI code generators that can write and test kernel code more efficiently than humans.
Nvidia's software for programming its AI chips, CUDA, has long been considered a highly sought-after skill set. However, as AI takes over the painstaking work of writing kernels, these engineers now spend their time setting goals, checking results, and stepping in when the AI gets stuck.
'In the past, we couldn't hire enough good-quality CUDA engineers, and now AI is filling the gap,' said Bing Xu, founder of the AI optimization startup INT21. According to Lightcast, there have been more US job postings requiring CUDA skills through the first eight months of 2026 than during all of 2025.
CUDA engineer jobs are evolving, with Nvidia remaining the largest employer hiring for these positions, advertising base salaries as high as $431,250 before stock and benefits. While AI lowers the barrier to entry for engineers with less experience, it simultaneously pushes veterans toward higher-level work, like overseeing AI.