AWS Trainium / NKI Kernel Expert
Core
Evaluate and improve kernel development tasks for AI workloads on AWS Trainium hardware, focusing on correctness, optimization, and migration from CUDA.
Role type
Senior IC AWS Trainium / NKI Kernel Expert
Builds
Optimized AI kernels and technical evaluations for AWS Trainium/Inferentia2
Domain
Cloud AI Infrastructure / Heterogeneous Computing
Deliverable
production ML models
Required skills
NKI kernel development, AWS Trainium architecture, CUDA to NKI migration, Tile-based computation, SBUF/PSUM/HBM memory management, DMA scheduling, Performance profiling, Numerical correctness analysis
Preferred skills
AWS Neuron SDK, Triton kernel development, NeuronCore-v2 architecture, FP32/BF16/FP8/INT8 workloads, nki.language, RLHF/rubric-based assessment
Technologies
AWS Trainium, Inferentia2, CUDA, Triton, Neuron Compiler, nki.language, XLA
Responsibilities
Review NKI kernel correctness and Trainium-specific patterns, Optimize Trainium performance and benchmark workloads, Analyze memory management across SBUF/PSUM/HBM, Evaluate CUDA to NKI migrations, Identify performance bottlenecks, Provide technical feedback on kernel implementations
Seniority
Senior, hands-on IC
