CUDA Math Libraries Intern

Job Description

Job Title: CUDA Math Libraries Intern - 2026. Location: Remote/Onsite - Warsaw Poland, Munich Germany, Zurich Switzerland

This role offers hands-on experience in designing accelerated mathematical primitives that are critical for AI and HPC applications; You will collaborate with senior engineers and gain exposure to parallel and GPU programming techniques while implementing and optimizing numerical algorithms.

Duties:

• Work with the team to understand software requirements and use cases.
• Analyze the performance of GPU and CPU implementations and identify optimization opportunities.
• Prototype and develop algorithms for single-node and multi-GPU cluster environments;
• Extend existing libraries or develop new modules used across AI, HPC, and quantum computing applications.
• Implement efficient numerical methods and address edge cases in computations;
• Participate in code reviews, testing, and documentation.

Qualifications:

• Currently pursuing an MS or PhD in Computational Science, Computer Science, Applied Mathematics, Engineering, or a related field.
• Strong programming skills in C/C++ and Python;
• Experience with parallel or GPU programming frameworks such as CUDA, OpenCL, OpenMP, MPI, or SHMEM.
• Familiarity with floating-point arithmetic, numerical error analysis, or hardware architectures.
• Knowledge of sparse or dense linear algebra algorithms is a plus;
• Experience with compiler optimizations, domain-specific language design, or sparse compilers is a bonus.

Benefits and Instructions:

• Mentorship from senior software engineers and hands-on exposure to GPU/HPC development.
• Opportunity to work on globally impactful projects in AI, HPC, and quantum computing;
• Competitive internship compensation and potential career growth opportunities at NVIDIA.
• Hybrid work setup available, combining on-site and remote collaboration;
• To apply, submit your resume in English and highlight relevant coursework or projects in GPU programming and numerical computing.