CUDA Engineer

A company is looking for a CUDA Engineer to engineer performant CUDA kernels with a focus on numerical correctness and efficiency.

Key Responsibilities

Engineer and optimize CUDA kernels for training and inference with large MoE models

Profile and debug CUDA applications to identify and resolve performance bottlenecks

Collaborate with team members to tackle technical challenges in building self-improving agents

Required Qualifications

Strong understanding of CUDA programming and performance optimization techniques

Familiarity with LoRA adapters and related optimizations such as Punica / S-LoRA kernels and Flash Attention

Experience in model training, particularly using reinforcement learning (RL)

Demonstrated ability to learn quickly and deliver impactful results

Prior experience in software development or engineering roles is preferred