Senior Staff AI Accelerator Performance Architect at Cerebras · Sunnyvale, CA · Joined
C
Senior Staff AI Accelerator Performance Architect
Cerebras · Sunnyvale, CA · Just now · 0 applicants
NewOn-siteFull-timeLead
Pay
$175k–$275kListed
Workplace
On-site
Job type
Full-time
Level
Lead
Source: jobright · Posted by joined-scrapper
About the role
Cerebras Systems is seeking a Senior Staff AI Accelerator Performance Architect to guide next-generation AI system evolution by developing performance models, analyzing workloads, and informing hardware and software roadmaps. The role requires deep expertise in performance analysis, architecture exploration, and quantitative validation for high-performance AI accelerator systems.
What you’ll do
Develop performance models for AI accelerator systems
Analyze workloads and identify bottlenecks
Evaluate architectural features and trade-offs
Inform hardware and software roadmaps
Profile workloads, form hypotheses, and validate with quantitative evidence
Communicate modeling assumptions, uncertainty, and recommendations clearly
What you’ll need
7+ years in performance analysis, modeling, or architecture exploration for CPUs, GPUs, or AI accelerators
Strong understanding of hardware architecture through hardware, compiler, kernel, or runtime work
Experience developing analytical, simulation-based, or trace-driven performance models using Python, C++ or similar
Solid understanding of processor architecture, memory systems, interconnects, parallel execution, and resource constraints
Ability to profile workloads and validate performance hypotheses with quantitative evidence
MS/PhD in Electrical Engineering, Computer Engineering, Computer Science or equivalent experience
Original job description
Cerebras Systems builds AI accelerator hardware and systems for high-speed AI training and inference. The Senior Staff AI Accelerator Performance Architect will guide next-generation AI system evolution by developing performance models, analyzing workloads and bottlenecks, evaluating architectural features, and informing hardware and software roadmaps.7+…