NVIDIA's AI Developer Tools organization is seeking to build the definitive benchmarks and evaluation frameworks for AI-powered CUDA programming and develop cutting-edge AI tools and methodologies for the future of accelerated computing.
Requirements
- Strong proficiency in Python and software engineering best practices
- Experience shipping production code or tools (beyond purely academic research)
- Experience with NVIDIA development tools (nvcc, CUDA toolkit)
- Strong analytical and problem-solving skills with attention to detail
- Experience with ML/AI experimentation workflows and evaluation methodologies
- Experience building or evaluating code generation models or AI-powered development tools
- Background with NVIDIA profiling and analysis tools (Nsight Compute, Nsight Systems) and/or the CUDA library ecosystem (cuDNN, cuBLAS, Thrust, CUB)
Responsibilities
- Design and build evaluation frameworks to assess AI models' ability to generate, optimize, and maintain CUDA code across the full software development lifecycle
- Develop benchmarks that represent real-world CUDA programming patterns and use cases across NVIDIA's ecosystem (kernels, libraries, multi-GPU applications)
- Contribute to cutting-edge AI research projects including novel training methodologies, tool development, and dataset curation initiatives
- Partner with teams developing CUDA-focused AI tools to provide evaluation insights, identify performance gaps, and integrate novel capabilities (e.g., RAG, profiling, web research)
- Create and curate high-quality datasets, leveraging both synthetic generation and real-world CUDA code to advance the state of AI-powered programming
- Explore and develop new AI tooling for developers, including IDE enhancements, cloud-served profiling services, and agent-ergonomic interfaces
- Conduct experiments to validate new approaches in areas like reinforcement learning for code optimization and multimodal representation learning
Other
- B.S. in Computer Science or related technical field or equivalent experience (M.S. preferred)
- 12+ years of relevant technical experience, with at least 5 years of hands-on CUDA programming experience (kernel development, optimization, debugging)
- Ability to work independently while collaborating effectively across teams
- Genuine interest in AI/ML and eagerness to learn new research methodologies
- Travel requirements not specified
