Intern - Memory Performance and Power Modeling

What you'll do

• Collect, process, and analyze memory traces from AI training and inference workloads.
• Develop workload traces, performance models, and architecture simulation environments; investigate system bottlenecks and architectural tradeoffs.
• Automate workload execution and analysis, validate models against measurements, and present findings in technical reports and architecture reviews.

What they ask for

• Pursuing an MS or PhD in Computer Engineering, Computer Science, Electrical Engineering, Data Science, or a related field.
• Coursework or research experience in computer architecture, machine learning/deep learning, parallel computing, operating systems, data structures and algorithms, or performance analysis.
• Programming experience in Python and C/C++.
• Strong analytical, problem-solving, and quantitative reasoning skills; excellent written and verbal communication skills.

Nice to have

• Experience with AI frameworks such as PyTorch, TensorFlow, JAX, or ONNX, or familiarity with AI models and workload profiling tools.
• Experience with simulation or architecture research; knowledge of accelerator architectures, memory systems, architecture simulators, or Linux development environments.

Pay

USD $50.24 - $55.82 an hour base salary

Tools

Python, C/C++, PyTorch, TensorFlow, JAX, ONNX, Gem5, Accel-Sim

AI focus

Analyze AI/ML training and inference workloads, including LLMs, recommendation systems, vision, multimodal, and generative AI applications, with a focus on memory traces, performance, and architecture.