PhD AI Systems & GPU Performance Intern

AMD; Santa Clara (CA); Hybrid; USD 55,000 - 69,000; Part time

Benefits offered by this job

AMD benefits at a glance

Job summary

AMD seeks a PhD candidate for an AI Systems & GPU Performance Internship. You will profile and optimize AI workloads on AMD Instinct GPUs, building reproducible benchmarks and automation while exploring techniques like quantization and mixed precision.

You will collaborate with researchers to evaluate performance across hardware/software environments and contribute data-driven recommendations for efficiency and scalability. This is an onsite/hybrid co-op through 40 hours per week.

Qualifications

• Pursuing a PhD in a technical field with expected graduation after internship.
• Hands-on programming experience in Python and C/C++, with accelerator programming exposure (CUDA, HIP, Triton).
• Experience with deep learning frameworks (PyTorch, JAX, TensorFlow) and large-scale model training/inference concepts.

Responsibilities

• Profile, benchmark, and optimize AI training and inference workloads on AMD Instinct GPUs.
• Develop reproducible benchmarking frameworks and automation for performance validation across stacks and platforms.
• Analyze end-to-end AI workflows and explore optimization techniques like quantization, mixed precision, and scheduling.

Skills

Python

C/C++

Linux

Benchmarking

Performance analysis

Education

PhD in Computer Science / Computer Engineering / Electrical Engineering / Applied Mathematics

Tools

CUDA

HIP

Triton

PyTorch

JAX

TensorFlow

rocProfiler

ROCm System Profiler

Omniperf

Nsight

Job description

AMD seeks a PhD candidate for an AI Systems & GPU Performance Internship. You will profile and optimize AI workloads on AMD Instinct GPUs, building reproducible benchmarks and automation while exploring techniques like quantization and mixed precision.

You will collaborate with researchers to evaluate performance across hardware/software environments and contribute data-driven recommendations for efficiency and scalability. This is an onsite/hybrid co-op through 40 hours per week.