Senior Software Engineer, CUTLASS Performance
Core
Benchmark deep learning model performance, identify GPU kernel optimization opportunities, and develop tooling to automate the performance optimization loop for the CUTLASS library.
Role type
Senior IC software performance engineer (GPU kernels)
Builds
Automated benchmarking and analysis tooling for CUTLASS kernel performance
Domain
High-performance computing, Deep Learning, GPU architecture
Deliverable
production ML models
Required skills
C++, Python, software performance analysis, GPU architecture knowledge, deep learning model architectures
Preferred skills
Performance benchmarking of DL frameworks (PyTorch, JAX, vLLM, TRT-LLM), performance regression systems development
Responsibilities
Benchmark inference and training passes of state-of-the-art deep learning models, Identify gaps between theoretical and realized performance, Develop tooling to automate the benchmarking and optimization loop, Serve as the authoritative resource on kernel performance and engage with cross-functional teams
Seniority
Senior, hands-on IC
