Be the Candidate They Can't Ignore.

Don't just send a resume; set yourself up so the right work finds you.

  • Look like a pro instantly with a stunning personal brand.

  • Gain instant "Top Ten" status so you aren't just another name in a pile.

  • Wake up to booked meetings instead of "rejection" emails.

Join the Club — Free Forever

Senior MLOps & LLM Performance Engineer (JAX/CUDA) - $90-$120/hr

  • Source: LinkedIn
  • Talent Wall
  • Remote
  • United Kingdom
  • Senior

Job Description

APPLY DIRECTLY HERE: https://t.mercor.com/Tdeld

Role Overview Join a leading AI lab’s cutting-edge GenAI team to build foundational AI models from the ground up. We’re seeking elite MLOps and ML Systems Engineers with hands-on expertise in LLM infrastructure (accelerator kernel programming, performance profiling, debugging distributed workloads, and high-throughput inference serving).

Please note: This is a 100% remote, full-time (40 hours/week) engagement open worldwide. It requires full dedication (no concurrent engagements/conflicts).

Key Responsibilities

  • Design extraordinarily challenging tasks across kernels (e.g., Pallas kernels on TPU), profiling, debugging, and inference serving to expose where frontier models fail on accelerator code.
  • Write accurate, well-structured solutions and evaluate AI-generated ML systems code, providing clear technical feedback that stands up to expert review.
  • Develop detailed rubrics covering kernel optimization, profiler output interpretation, distributed systems reasoning, and serving latency trade-offs.

Core Qualifications

  • Experience: 2+ years of highly technical, hands-on professional experience in ML systems, ML infrastructure, model serving, or accelerator performance engineering. (This is a systems role, not an applied data science role).
  • Hardware & Frameworks: Deep working production experience with JAX and/or PyTorch. Experience with custom operators, sharding (FSDP, Megatron), or graph-level work (XLA, MLIR) is a strong plus.
  • Accelerators & Kernels: Hands-on experience writing/optimizing custom kernels (Pallas, Triton, CUDA, CUTLASS) and working with modern accelerators (TPU v4-v6e, H100, MI300).
  • Inference & Serving: Practical experience serving LLMs at scale using frameworks like vLLM, SGLang, or TensorRT-LLM (continuous batching, paged attention).
  • Availability: Must be able to commit to 40 hours per week exclusively.

Engagement Details & Compensation

  • Work Structure: Fully remote independent contractor role (managed via Cincinnatus LLC or a local EOR depending on your country). Flexible schedule within the 40-hour requirement.
  • Compensation: Highly competitive flat rate of $90.00 to $120.00 per hour .
  • Payments: Processed securely and weekly via Stripe or Wise based on services rendered.

(We consider all qualified applicants without regard to legally protected characteristics and provide reasonable accommodations upon request.)

Skills

  • cuda
  • PyTorch
  • microsoft-graph
  • triton
  • vllm1
  • sglang
  • stripe

Create Free Account & Apply
  • Free personal website
  • Built-in Talent Agent
  • Apply to future jobs in 1 click
  • Get matched to similar roles
or
Apply as a guest this time

You'll apply successfully, but won't be able to track your application, save this job, or get matched to similar roles.

This listing is a publicly available opportunity curated by Try Club to help talent discover relevant roles. Try Club is an independent platform and is not affiliated with Talent Wall

Get Similar Jobs

Browse more similar roles on TRY and find your perfect match.

Show Me Similar Jobs

Stop Searching. Start Thriving.

Why waste time chasing leads when the perfect projects can find you?
With TRY, clients actively searching for professionals with your skills can discover you in seconds. Whether it's a quick gig or your dream role, we connect you with opportunities that matter.

  • Get found faster
  • Let projects come to you
  • Build a career on your terms

Take control of your career today.

Claim Free Website in 2 Minutes