Lead Full Stack Machine Learning Engineer

Cerebras Cerebras · Semiconductors · India · Software

This role focuses on bringing up and optimizing open-source AI models and frameworks on Cerebras' wafer-scale hardware. It involves working across the full software stack, from model translation and compiler optimizations to runtime integration and performance tuning, with a strong emphasis on debugging and improving the bring-up process for future models.

What you'd actually do

  1. Contribute to the end-to-end bring up of frameworks for RL, inference serving, ML models on Cerebras CSX systems.
  2. Work across the stack: model architecture translation, graph lowering, compiler optimizations, runtime integration, and performance tuning.
  3. Debug performance and correctness issues spanning model code, compiler IRs, runtime behavior, and hardware utilization.
  4. Propose and prototype improvements across tools, APIs, or automation flows to accelerate future bring ups.

Skills

Required

  • Bachelor’s, Master’s, or PhD in Computer Science, Engineering, or a related field
  • Comfort navigating the full AI toolchain: Python modelling code, compiler IRs, performance profiling, etc.
  • Strong debugging skills across performance, numerical accuracy, and runtime integration.
  • Experience with deep learning frameworks (e.g., PyTorch, TensorFlow) and familiarity with model internals (e.g., attention, MoE, diffusion).
  • Proficiency in C/C++ programming and experience with low-level optimization.
  • Strong background in optimization techniques, particularly those involving NP-hard problems.

Nice to have

  • RL frameworks

What the JD emphasized

  • 10+ years’ experience
  • full AI toolchain
  • performance profiling
  • low-level optimization
  • optimization techniques
  • NP-hard problems

Other signals

  • Bring up state-of-the-art open-source models, frameworks and data engineering
  • Work across the stack: model architecture translation, graph lowering, compiler optimizations, runtime integration, and performance tuning
  • Debug performance and correctness issues spanning model code, compiler IRs, runtime behavior, and hardware utilization
  • Propose and prototype improvements across tools, APIs, or automation flows to accelerate future bring ups