ML Systems Performance Engineer

Cerebras Systems builds the world's largest AI chip, 56 times larger than GPUs. This architecture allows Cerebras to deliver industry-leading training and inference speeds; over 10 times faster than GPU-based hyperscale cloud inference services. This order of magnitude increase in speed is transforming the user experience of AI applications, unlocking real-time iteration and increasing intelligence via additional agentic computation. Cerebras works with the leading model labs, global enterprises, and cutting-edge AI-native startups. OpenAI recently announced a multi-year partnership https://openai.com/index/cerebras-partnership/ with Cerebras, to deploy 750 megawatts of scale, transforming key workloads with ultra high-speed inference. ABOUT THE ROLE Engineers on the inference performance team operate at the intersection of hardware and software, driving end-to-end model inference speed and throughput. Their work spans low-level kernel performance debugging and optimization, system-level performance analysis, performance modeling and estimation, and the development of tooling for performance projection and diagnostics. RESPONSIBILITIES - Build performance models (kernel-level, end-to-end) to estimate the performance of state of the art and customer ML models. - Optimize and debug our kernel micro code and compiler algorithms to elevate ML model inference speed, throughput and compute utilization on the Cerebras WSE. - Debug and understand runtime performance on the system and cluster. - Develop tools and infrastructure to help visualize performance data collected from the Wafer Scale Engine and our compute cluster. REQUIREMENTS - Bachelors / Masters / PhD in Electrical Engineering or Computer Science. - Strong background in computer architecture. - Exposure to and understanding of low-level deep learning / LLM math. - Strong analytical and problem-solving mindset. - 3+ years of experience in a relevant domain (Computer Architecture, CPU/GPU Performance, Kernel Optim

Department: Software; Team: Software