NKI Kernel Optimization Engineer for Trainium AI

Mercor; San Francisco (CA); On-site; USD 120,000 - 180,000; Full time

Job summary

Mercor is seeking a specialist to evaluate the quality, correctness, and hardware-appropriateness of Neuron Kernel Interface (NKI) development tasks used to train and evaluate a frontier AI lab's models. You will assess CUDA→NKI migration fidelity, Trainium-specific performance, cross-platform numerical-correctness standards, and provide rubric-based written feedback.

Reporting into Mercor's AI systems group, you will work with kernel engineers and ML developers to ensure robust hardware-aware

Qualifications

• 2+ years hands-on kernel development with NKI targeting Trainium/Inferentia2.
• Expertise in NKI patterns: tile-based computation, SBUF/PSUM/HBM memory hierarchy, DMA.
• Experience assessing CUDA→NKI migration quality and Trainium profiling.

Responsibilities

• Evaluate quality, correctness, and hardware-appropriateness of NKI development tasks.
• Provide rubric-based written feedback on migration fidelity and performance.

Skills

NKI kernel development

CUDA to NKI migration

Trainium hardware

Tile-based computation

Memory hierarchy management

DMA orchestration

Performance profiling

Numerical correctness standards

Tools

AWS Neuron SDK

Neuron Compiler

NKI libraries

Job description

Mercor is seeking a specialist to evaluate the quality, correctness, and hardware-appropriateness of Neuron Kernel Interface (NKI) development tasks used to train and evaluate a frontier AI lab's models. You will assess CUDA→NKI migration fidelity, Trainium-specific performance, cross-platform numerical-correctness standards, and provide rubric-based written feedback.

Reporting into Mercor's AI systems group, you will work with kernel engineers and ML developers to ensure robust hardware-aware