All AI training jobs

Mercor · Finance & specialist

Trainium (NKI) Kernel Expert - review AI outputs in your specialty

Listed on Mercor as “Trainium (NKI) Kernel Expert”

$70-$90/hrRemoteContractPaid in USD
ShareWhatsAppTelegramEmail

How this rate compares

This listing

$70–$90 per hour

Typical in Finance & specialist

$60–$100 per hour

Within the typical range

Typical: the middle of the lowest and highest rates posted by 345 listings in this category on Seoulstart, in US dollars.

What this actually is

You bring your specialist expertise to AI evaluation. The shape of the work varies but the pattern is the same: review outputs, rate quality, write prompts, flag errors. The platform title (Trainium (NKI) Kernel Expert) reflects the rate band and the expertise required, not the day-to-day work.

How Mercor picks people

  1. Upload your CV and answer the role's written questions.
  2. Do a video interview with Mercor's AI interviewer. Most take about 20 minutes, in English unless the role says otherwise.
  3. Some roles add a skills test. A test you pass once counts for similar roles.
  4. The project team reviews your results and sends offers to the people it picks.

Before you start

  • Take Mercor's free practice interview first. It does not count toward any application.
  • Use Chrome on a computer with a camera and microphone, in a quiet room.
  • You can retake an official interview up to 3 times, but hiring teams see how many times you retook it.

Not every applicant gets an offer. Prepare well for the interview.

How to prepare for the Mercor interview

Source: Mercor's help pages. talent.docs.mercor.com

Can you do this on your visa?

F-2 / F-4 / F-5 / F-6: open. E-1 to E-7: needs concurrent-employment permit. D-2 / D-4 students: S-3 permit, 20 hr/week cap. D-10 / D-8: case by case.

Korean tax on USD income

First 5 years in Korea: foreign-source income only taxed if remitted into Korea. After year 5: worldwide income. Full tax guide.

Original posting from Mercor

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'll assess CUDA→NKI migration fidelity, Trainium-specific performance-optimization quality, and cross-platform numerical-correctness standards - and provide clear, rubric-based written feedback.

Basic Qualifications

• 2+ years of hands-on experience developing or optimizing kernels using the Neuron Kernel Interface (NKI) targeting AWS Trainium/Inferentia2 hardware

• Strong understanding of NKI-specific development patterns: tile-based computation, SBUF/PSUM/HBM memory-hierarchy management, partition-dimension constraints, and DMA orchestration

• Demonstrated experience assessing CUDA→NKI migration quality

• Familiarity with Trainium-specific performance profiling (NeuronCore pipeline utilization, tensor-engine throughput, memory-bandwidth bottlenecks)

• Experience defining or evaluating cross-platform numerical-correctness standards (GPU vs Trainium accumulation order, rounding behavior, mixed-precision semantics)

Preferred Qualifications

• Direct 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 (NeuronCore-v2 architecture, on-chip SRAM topology, supported data types: FP32/BF16/FP8/INT8)

• Experience benchmarking ML training workloads on Trn1/Trn2 instances

Quoted from Mercor's public listing on 2026-11-13. We don't edit platform copy; honest framing is in the title and the “what this actually is” block above.

Mercor pays MantaWork a fee if it hires you through this link. Your pay is the same.