AWS Trainium / NKI Kernel Expert

About the role

Anyone AI is seeking an experienced AWS Trainium / Neuron Kernel Interface (NKI) engineer to join a part‑time, project‑based consulting engagement. You will evaluate and improve kernel development tasks for AI workloads, ensuring implementations are both correct and optimized for Trainium hardware.

Key responsibilities

• Review NKI kernel correctness and Trainium‑specific development patterns.
• Migrate CUDA kernels to NKI, assessing technical fidelity.
• Optimize performance, benchmark workloads, and identify bottlenecks on Trainium.
• Manage memory across SBUF, PSUM, and HBM, and orchestrate DMA scheduling.
• Provide detailed technical feedback and quality assessments of kernel implementations.

Required profile

• 2+ years of hands‑on experience with NKI kernel development or optimization.
• Strong knowledge of AWS Trainium and/or Inferentia2 architectures.
• Proven ability to evaluate CUDA→NKI migrations and profile AI workloads.
• Excellent written communication for technical feedback.

Required skills

• AWS Trainium
• Inferentia2
• Neuron Kernel Interface (NKI)
• CUDA
• Triton
• AWS Neuron SDK / Neuron Compiler
• Tile‑based computation, SBUF/PSUM/HBM memory hierarchy
• DMA orchestration
• FP32, BF16, FP8, INT8 workloads
• Profiling and benchmarking on Trn1/Trn2 instances
• XLA custom calls, nki.language, @nki.jit

What we offer

• Remote work flexibility.
• Project‑based consulting contract.
• Opportunity to influence cutting‑edge AI kernel optimization.