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

Listed on Mercor ↗

mercor.com

ApplyApply

About the Role

This is a fully remote role open to candidates worldwide. Depending on where you are based, your engagement will be arranged through Cincinnatus LLC or one of its trusted employer-of-record partners in your country, with the opportunity to be placed at a leading AI lab as part of their extended workforce. This is a 40-hour full-time engagement, with no conflicts/no other engagements. Cincinnatus LLC is an enterprise staffing company that partners with leading technology companies to source and engage highly skilled professionals for contingent and contract-based opportunities.

What You'll Do

• Design challenging tasks across kernels, profiling, debugging and inference serving, including adversarial problems that expose where frontier models fail on accelerator code (for example Pallas kernels on TPU), and write accurate, well-structured solutions to them.
• Guide research and engineering teams to close knowledge gaps and improve AI model performance on ML systems, compilers and training infrastructure topics.
• Evaluate ML systems tasks and solutions, and provide clear written technical feedback that stands up to expert review.
• Develop guidelines and detailed rubrics covering kernel optimization, profiler output interpretation, distributed systems reasoning, and serving throughput and latency trade-offs.
• Collaborate with other subject matter experts to keep training data consistent and accurate.

You're a Good Fit If You

• 2+ years of hands-on professional experience in ML systems, ML infrastructure, model serving, or accelerator performance engineering. This is a hands-on systems role rather than an applied modelling or data science one.
• Practical experience in at least one of the following, with more than one a strong plus: writing or optimizing custom kernels (Pallas, Triton, CUDA, CUTLASS); performance profiling and trace analysis (JAX/XLA profiler, Kineto, torch.profiler, Nsight);
• Working production experience with JAX and/or PyTorch. Framework-level depth is a strong plus: custom operators, sharding and distributed training (pjit/shard_map, FSDP, Megatron, MaxText), or compiler and graph-level work (XLA/StableHLO, torch.compile, MLIR).
• Experience with modern accelerators such as TPU (v4 to v6e), H100, B200, Trainium or MI300, and the ability to reason about throughput, latency and memory trade-offs.
• Demonstrable career progression.
• Ability to engage reliably for at least 40 hours/week during weekdays.
• Strong written communication skills and the ability to explain complex technical decisions clearly.

Role Highlights

Pay & Payout

Independent contractor engagement. Apply via the link below - Mercor handles onboarding and pays weekly via Stripe or Wise.

Common Questions

Opens Mercor in a new tab. You pay nothing; we may earn a referral fee.