ML Systems Performance Engineer — Hardware Co-Design

Cerebras Systems; Sunnyvale (CA); On-site; USD 180,000 - 240,000; Full time

Job summary

Cerebras Systems in Sunnyvale, CA is seeking an engineer for the Core ML team to map novel ML algorithms to our Cerebras architecture and benchmark their efficiency.

You will develop analytical models, run end-to-end profiling, and collaborate with kernel, compiler, and hardware teams to push performance boundaries.

Strong background in architecture, ML systems, and Python/C++ is required; experience with roofline analysis and CUDA is a plus.

Qualifications

• Advanced degree or equivalent practical experience in a related field.
• Strong knowledge of computer architecture and systems performance.
• Experience with analytical performance modeling and benchmarking.
• Proficiency in Python and C++ programming.
• Ability to analyze and optimize ML training/inference workloads.
• Experience profiling ML/HPC systems across hardware platforms.

Responsibilities

• Build analytical models and benchmarks for state‑of‑the‑art ML training and inference.
• Characterize asymptotic behavior and how tradeoffs vary with size and scale.
• Construct Pareto frontiers across quality, latency, throughput, and memory.
• Develop prototypes and benchmarks for Cerebras WSE and baselines.
• Analyze system bottlenecks across kernel, compiler, runtime, and communication.
• Evaluate techniques like parallel token generation, attention, sparsity, and MoE.
• Collaborate with researchers and teams to guide implementation directions.
• Develop tools/visualizations to communicate performance projections and tradeoffs.
• Present conclusions and recommendations through technical reports and reviews.

Skills

Computer architecture

Parallel computing

Systems performance

ML systems fundamentals

Performance modeling

Benchmarking

System simulation

Python

C++

Profiling & debugging ML/HPC systems

Education

Bachelor’s / Master’s / PhD or equivalent

Tools

CUDA

Triton

PyTorch

JAX

Open-source ML systems

Job description

Cerebras Systems in Sunnyvale, CA is seeking an engineer for the Core ML team to map novel ML algorithms to our Cerebras architecture and benchmark their efficiency.

You will develop analytical models, run end-to-end profiling, and collaborate with kernel, compiler, and hardware teams to push performance boundaries.

Strong background in architecture, ML systems, and Python/C++ is required; experience with roofline analysis and CUDA is a plus.