Job description
About the job
Mercor connects elite creative and technical talent with leading AI research labs. Headquartered in San Francisco, our investors include Benchmark, General Catalyst, Peter Thiel, Adam D'Angelo, Larry Summers, and Jack Dorsey.
Position: Trainium (NKI) Kernel Expert
Type:Contract
Compensation:$70–$90/hour
Location:Remote
Role Responsibilities
Evaluate the quality, correctness, and hardware-appropriateness of Neuron Kernel Interface (NKI) development tasks for training and evaluating AI models.
Assess CUDA→NKI migration fidelity and Trainium-specific performance-optimization quality.
Review cross-platform numerical-correctness standards and provide clear, rubric-based feedback.
Collaborate with AI research teams to ensure consistency and coverage across datasets.
Work independently and asynchronously to improve AI model performance.
Qualifications
Must-Have
2+ years of experience developing or optimizing kernels using NKI targeting AWS Trainium/Inferentia2 hardware.
Strong understanding of NKI-specific development patterns and memory-hierarchy management.
Experience assessing CUDA→NKI migration quality.
Familiarity with Trainium-specific performance profiling.
Experience defining or evaluating cross-platform numerical-correctness standards.
Preferred
Experience with AWS Neuron SDK, Neuron Compiler internals, or contributions to NKI kernel libraries.
Prior CUDA or Triton kernel development.
Familiarity with Trainium hardware specifications.
Experience benchmarking ML training workloads on Trn1/Trn2 instances.
Application Process (Takes 20–30 mins to complete)
Upload resume
AI interview based on your resume
Submit form
Resources & Support
For details about the interview process and platform information, please check:
For any help or support, reach out to:
PS: Our team reviews applications daily. Please complete your AI interview and application steps to be considered for this opportunity.
Originally posted on Himalayas
Who can apply
Eligible countries: United States. Accepted UTC offsets: UTC-10, UTC-9, UTC-8, UTC-7, UTC-6, UTC-5, UTC+14. Review the full description for employer-specific work authorization, residency and schedule requirements.