We are seeking a Large-Scale Performance Modeling Engineer Intern to help design and extend performance models that predict system behavior at cluster scales beyond what is physically available for testing. In this role, you’ll work at the intersection of systems architecture, networking, and AI infrastructure, contributing to forward-looking models that influence future platform design. This internship offers hands-on exposure to large-scale compute systems, simulation tools, and real-world performance analysis challenges.

You may collaborate with performance modeling frameworks and network simulation toolchains (such as ATLAHS) to explore how compute, memory, and networking scale in next-generation clusters.

• Develop and enhance performance models to predict behavior of large-scale compute clusters beyond current hardware limits
• Analyze system-level performance across compute, memory, and networking components
• Integrate and evaluate modeling approaches using network simulators and toolchains (e.g., ATLAHS)
• Automate performance experiments, data collection, and analysis using scripts and tooling
• Document findings and clearly communicate results, assumptions, and trade-offs to engineering stakeholders
• Collaborate with architects and engineers to validate models and refine simulation methodologies

Basic Qualifications:

• Currently pursuing a degree in Computer Science, Computer Engineering, Electrical Engineering, or a related technical field
• Foundational understanding of:

· x86 and/or ARM server architectures

· High-speed networking (Ethernet)

· Linux-based operating systems

• Experience with Python or similar scripting languages for automation, testing, and data analysis
• Ability to clearly explain complex technical concepts through written reports and verbal discussions
• Demonstrated ability to quickly learn new technologies, including hardware protocols and software stacks

Preferred Qualifications:

• Fluent in both English and Mandarin Chinese
• Familiarity with one or more of the following:

· Heterogeneous computing environments

· GPU programming or software stacks (e.g., NVIDIA CUDA, AMD ROCm)

· Collective communication libraries (e.g., NVIDIA NCCL, AMD RCCL)

· Distributed AI training frameworks or middleware (e.g., PyTorch Distributed, DeepSpeed ZeRO)

· Switch or network operating systems (e.g., SONiC)