We are open for seniors, mids, and juniors (of course, then with lower salary), but at least first experience with GPU performance and kernel optimization is must have.

Remote

• Explore and analyze performance bottlenecks in ML training and inference.
• Develop and optimize high-performance computing kernels in Triton, CUDA, and/or ROCm.
• Implement programming solutions in C/C++ and Python.
• Deep dive into GPU performance optimizations to maximize efficiency and speed.
• Collaborate with the team to extend and improve existing machine learning compilers or frameworks such as MLIR, Pytorch, Tensorflow, ONNX Runtime, TensorRT. (This is optional but beneficial)

• Strong programming skills in C/C++ and Python.
• Deep understanding and experience in GPU performance optimizations.
• Proven experience with kernel optimizations on CUDA, ROCm, or other accelerators.
• General experience with the training and deployment of ML models
• Experience with distributed systems development or distributed ML workloads
• Bachelor's, Masters or PhDs degree in Computer Science, Electrical Engineering, or a related field.

• Experience with innovative OSS projects like FlashAttention, mlc-llm, vllm, SGLang.
• Experience with machine learning compilers or frameworks such as TVM, MLIR, Pytorch, Tensorflow, ONNX Runtime, TensorRT.