Skip to content
Software & codeContractSome African countries

MLOps Engineer, LLM Performance (Kernels, Profiling, Serving)

On Mercor

$90–$120/hr

ML systems engineers write and assess hard training tasks on accelerator kernels, profiling, debugging and LLM inference serving for a leading AI lab's generative AI team. It is a full-time remote engagement open worldwide, suited to engineers with production JAX or PyTorch experience.

Referral link: ASV may earn a bonus if you are accepted, at no cost to you. Disclosure

Pay

$90–$120/hr

US dollars, as the listing states it.

Type
Contract
Hours
Full-time: at least 40 hours a week on weekdays, with no other engagements
Africa
Some countries
Platform
Mercor
Posted
Oct 22, 2026

01 · The role

What you would do

  • Create hard problems in four areas (kernels, profiling, debugging and serving), some built to trip up frontier models on accelerator code, with correct worked solutions
  • Review ML systems tasks and answers, writing technical feedback that holds up to expert scrutiny
  • Draft rubrics and guidelines for judging kernel tuning, profiler readings, distributed-systems reasoning and the balance of serving throughput against latency
  • Coordinate with fellow specialists so the training data stays accurate and consistent

02 · Requirements

What they ask for

  • Two or more years of hands-on work in ML systems or infrastructure, model serving or accelerator performance
  • Practical experience in custom kernels (Pallas, Triton, CUDA or CUTLASS), profiling and trace analysis, debugging distributed or accelerator-bound workloads, or serving LLMs at scale
  • Production experience with JAX or PyTorch, and with modern accelerators such as TPUs, H100, B200, Trainium or MI300
  • Availability for at least 40 hours a week on weekdays
  • Strong written communication

Helpful, not required

  • Experience with JAX, XLA or TPUs
  • Framework-level depth, such as custom operators, sharding and distributed training, or compiler work
  • MLOps
  • Kernel optimization
  • Performance profiling
  • LLM inference
  • JAX
  • PyTorch

03 · Eligibility

Who can apply

Remote, and the listing accepts candidates in any country, contracted through Cincinnatus LLC or a local employer-of-record partner. Mercor's payment countries in October 2026 did not include Kenya, so applicants in Kenya should confirm before applying.

Some African countries

Open in some African countries only: check the listing for which. Mercor in Africa

Apply from where you live, as yourself. Using a VPN or someone else's identity to get around a country rule breaks these platforms' terms, and it is the quickest way to lose an account and any unpaid earnings.

04 · Applying

How to get in

You apply through Mercor's own process. The steps from sign-up to first paid task are on its profile.

05 · Source

About this listing

We last confirmed this opening on Oct 30, 2026, and it was posted on Oct 22, 2026. Rates and hours come from the listing unless the note beside the rate says otherwise. Where the listing names no countries, the Africa status is our reading of it and of what is known of Mercor's country rules, set out on its profile, and never marks a role more open than they show. The listing itself has the final word.

The listing on work.mercor.com(opens in a new tab)

More open roles like this.

All openings

The platform

Specialist AI-training contracts for frontier labs, paid weekly via Stripe.

$20–$200/hrHighly selectiveGeneralist, Software & code, Math & science, Domain experts, Creative

Mercor

Ready to apply? Read the listing first.

Check the rate, the hours and who can apply against the listing, follow the platform's own steps, and keep a second application going.

MLOps Engineer, LLM Performance (Kernels, Profiling, Serving)

$90–$120/hr

Apply(opens in a new tab)