onsite
System Software Engineer - CUDA Chips - NVIDIA
Software Engineer
Lead the development of NVIDIA’s CUDA driver, enhancing GPU acceleration for deep learning, scientific computing, autonomous driving, gaming, and VR. Drive kernel‑mode driver improvements, hardware interaction, and performance optimizations on Linux platforms.
About the role
Key Responsibilities
- Design, implement, and maintain the CUDA driver stack, ensuring robust interaction between GPU hardware, kernel mode drivers, and the operating system.
- Collaborate with hardware teams to expose new GPU features and optimize driver performance for diverse workloads such as deep learning, scientific simulation, autonomous driving, gaming, and virtual reality.
- Debug and resolve complex kernel‑level issues, leveraging low‑level profiling and tracing tools to identify bottlenecks and stability problems.
- Contribute to driver architecture, code reviews, and documentation, ensuring high code quality and maintainability.
- Participate in cross‑functional teams to define feature roadmaps and deliver incremental improvements that maximize GPU utilization.
Requirements
- Strong experience in C/C++ system programming and kernel development on Linux.
- Deep understanding of GPU architecture, driver design, and CUDA programming model.
- Proficiency with low‑level debugging, profiling, and performance tuning tools.
- Excellent problem‑solving skills and ability to work in a fast‑paced, collaborative environment.
- Experience with deep learning or high‑performance computing workloads is a plus.