Director / Sr. Manager, AI Inference Model Scaling

About the role

Cerebras Systems builds the world's largest AI chip, which delivers training and inference speeds over 10 times faster than GPU-based cloud services. The Inference Model Scaling team enables foundation models and generative AI workloads to run efficiently on the Wafer-Scale Engine, covering the compiler frontend, model transformation pipeline, graph optimization infrastructure, high-performance kernels and runtime integration. Cerebras is looking for an experienced engineering leader to build and scale this organization, defining its technical vision, organizational strategy and roadmap for a globally distributed team.

Key responsibilities

• Define the technical roadmap and strategy, set technical direction across teams and lead design reviews and engineering standards.
• Hire, mentor and grow the engineering team, develop future technical leaders and drive headcount and investment priorities.
• Partner with Cloud Platform, ML and Hardware teams, work with Product Management and collaborate with customers and solution architects on model bring-up.
• Influence hardware/software co-design through model enablement insights and balance rapid model support with long-term compiler architecture.
• Own planning and execution across multiple initiatives and drive predictable delivery for strategic customer commitments.

Required profile

• BS, MS or PhD in Computer Science, Computer Engineering or a related field.
• 12+ years building compiler, ML systems or infrastructure software.
• 5+ years leading engineering teams.
• Deep experience with modern compiler infrastructure and strong understanding of graph compilation and optimization.
• Experience delivering production-quality software and strong cross-functional leadership.

Required skills

• Modern compiler infrastructure (LLVM, MLIR, XLA, TVM, Torch FX or similar)
• Python and C++
• Preferred: compiler frontends for AI accelerators, PyTorch, JAX, TensorFlow or ONNX, LLM inference or training systems, distributed compilation, hardware architect collaboration, leading teams through rapid growth