MLOps Engineer: GPU Kernels, Profiling & Inference

SaidGig; New York (NY); On-site; USD 124,000 - 165,000; Full time

Job summary

SaidGig is seeking a hands-on ML systems engineer to help develop foundational large language models. The role focuses on GPU kernels, performance profiling, trace analysis, distributed workload debugging, and high-throughput inference serving.

You will design kernel optimization tasks, provide detailed feedback, support research teams, and create evaluation rubrics. Collaboration with experts will maintain accurate training data and advance model performance across training infrastructure and

Qualifications

• 2+ years hands-on experience in ML systems or ML infrastructure.
• Experience with CUDA, Triton, or Pallas for GPU kernel development.
• Production experience with JAX and/or PyTorch.
• Familiarity with accelerators (A100/H100/TPU) and throughput vs latency trade-offs.

Responsibilities

• Design ML systems tasks in GPU kernels, profiling, debugging, and inference serving, then produce accurate, well-structured solutions.
• Evaluate technical tasks and solutions, providing clear written feedback that can withstand detailed review.
• Support research and engineering teams in closing knowledge gaps and improving model performance across ML systems, training infrastructure, and framework-level subjects.
• Create detailed guidelines and evaluation rubrics for kernel optimization, profiler-output interpretation, distributed-systems reasoning, and serving throughput and latency trade-offs.
• Partner with subject matter experts to maintain consistent, accurate training data.

Skills

ML systems experience

GPU kernel development

Profiling & trace analysis

PyTorch/JAX

Tools

CUDA

Triton

Pallas

Kineto

torch.profiler

Job description

SaidGig is seeking a hands-on ML systems engineer to help develop foundational large language models. The role focuses on GPU kernels, performance profiling, trace analysis, distributed workload debugging, and high-throughput inference serving.

You will design kernel optimization tasks, provide detailed feedback, support research teams, and create evaluation rubrics. Collaboration with experts will maintain accurate training data and advance model performance across training infrastructure and