Performance Software Intern, Deep Learning Libraries - 2027
Core
Developing optimized code to accelerate linear algebra and deep learning operations on NVIDIA GPUs, focusing on code lower in the deep learning software stack down to the GPU HW.
Role type
Performance Software Intern (Deep Learning Libraries)
Builds
Optimized compute kernels for core deep learning operations (matrix multiplies, MoE, Attention) for NVIDIA GPUs
Domain
AI Infrastructure / High-Performance Computing / GPU Architecture
Deliverable
production ML models
Required skills
Parallel programming, Performance analysis, Assembly programming, Computer architecture, Debugging, Test design, Linear algebra, Numerical methods
Preferred skills
CUDA GPU programming, LLVM, TVM tensor expressions, TensorFlow MLIR
Responsibilities
Writing highly tuned compute kernels for deep learning operations, Collaborating with compiler, training/inference, and hardware teams, Identifying bottlenecks and optimizing resource utilization
Seniority
Intern
