Performance Software Intern, Deep Learning Libraries - 2027
Salary not listed • China, Shanghai • Internship
Posted 1mo ago
Remote Work Policy
On-site
Employment Type
Internship
Categories
Machine Learning Engineer
About the job
NVIDIA is seeking a Performance Software Intern for Deep Learning Libraries. This role involves developing optimized code to accelerate linear algebra and deep learning operations on NVIDIA GPUs, contributing to the foundational software that powers AI advancements. The intern will focus on achieving peak GPU efficiency for current and future GPU generations, working with code at a low level, close to the GPU hardware. The team's work is exemplified by the open-source CUTLASS project, which demonstrates performant matrix multiplication on NVIDIA's Tensor Cores using CUDA.
Responsibilities
- Write highly tuned compute kernels for core deep learning operations like matrix multiplies, MoE, and Attention.
- Adhere to software engineering best practices, including regression testing and CI/CD.
- Collaborate with compiler teams on generating optimal assembly code.
- Work with deep learning training and inference performance teams to identify layers needing optimization.
- Engage with hardware and architecture teams on programming models for new deep learning hardware features.
- Identify performance bottlenecks, optimize resource utilization, and improve throughput.
Requirements
- Currently pursuing a Masters or PhD degree in Computer Science, Computer Engineering, Applied Math, or a related field.
- Demonstrated strong programming and software design skills, including debugging, performance analysis, and test design.
- Experience with performance-oriented parallel programming (e.g., OpenMP, pthreads), even if not specifically on GPUs.
- Solid understanding of computer architecture.
- Some experience with assembly programming.