THE ROLE:
AMDis looking for aspecializedsoftware engineer who is passionate about improving the performance of key applications and benchmarks.You will be a member of a core team ofincredibly talentedindustry specialists and will work with the very latest hardware and software technology.
THE PERSON:
The ideal candidate should be passionate about software engineering and possess leadership skills to drive sophisticated issues to resolution. Able to communicate effectively and work optimally with different teams across AMD.
KEY RESPONSIBILITIES:
- Design, develop, and optimize algorithms for collective communication operations (e.g., All-Reduce, All-to-All, Broadcast) within AMDs RCCL.
- Analyze and tune the performance of collective communication libraries on large-scale GPU clusters, focusing on latency, bandwidth, and scalability over high-speed network fabrics.
- Integrate and validate RCCL with various network transport layers and protocols, such as UEC, RoCE (RDMA over Converged Ethernet), and custom interconnects.
- Collaborate closely with hardware, driver, and machine learning framework teams to co-design and debug system-level performance issues.
- Develop robust benchmarking and profiling tools to identify and resolve bottlenecks in the communication software stack.
- Contribute to the upstream open-source RCCL project and stay current with the latest advancements in the field.
- Provide expert guidance on GPU cluster network topology and configuration to maximize collective communication performance.
PREFERRED EXPERIENCE:
- 10+ years of experience in software development with a strong focus on high-performance computing or distributed systems.
- Highly proficient in C/C++ programming and debugging in a Linux environment.
- Experience with performance analysis, profiling, and debugging of complex, distributed systems in a Linux environment.
- Proven track record of optimizing software for specific hardware architectures.
- Strong analytical and problem-solving skills, with a proven ability to diagnose and resolve complex performance issues.
- Hands-on experience with AMD ROCm RCCL or similar GPU collective communication libraries (NVIDIA NCCL, MSICCL,oneCCL, Open MPI, MPICH, or other MPI implementationsetc.), RoCEv2/RDMA is a huge plus.
- Effective communicationand problem-solving skills.
ACADEMIC CREDENTIALS:
- Bachelor s orMastersdegree in Computer Science, Computer Engineering, Electrical Engineering, or equivalent
