ML Systems Engineer - High-Performance GPU & Distributed Training

Digitailor; Menlo Park (CA); On-site; USD 180,000 - 230,000; Full time

Job summary

Digitailor in Menlo Park, CA, is building high-performance AI/ML training and inference systems for image, video, and world-model workloads. You will optimize performance across GPU kernels, distributed execution, and multi-node serving.

The role emphasizes low-level GPU optimization with CUDA and Triton, designing distributed engines for diffusion models, and improving inter-GPU communication with NCCL, RDMA, InfiniBand, or RoCE.

Qualifications

• 2+ years in deep learning training or inference systems, distributed systems, or related ML systems work.
• 1+ year hands-on ML systems or GPU performance experience in a current or recent role.
• Experience authoring core features in an inference or training framework (e.g., vLLM, SGLang, TensorRT-LLM, Megatron).
• Hands-on experience writing CUDA, CUTLASS, Triton, or PTX kernels for NVIDIA GPUs.
• Strong fundamentals in ML and CS, with a CS-related degree.
• Experience optimizing inference or training performance (GPU kernels, runtimes, distributed execution).
• Experience with diffusion, video, image, or multimodal workloads is valuable.

Responsibilities

• Optimize training and inference performance for image, video, and world-model workloads.
• Use Nsight and related tools to identify bottlenecks across kernels, memory, systems, and clusters.
• Implement low-level GPU optimizations with CUDA and Triton.
• Design and improve distributed training and inference engines for diffusion models.
• Improve communication performance across GPUs and nodes using NCCL, RDMA, InfiniBand, or RoCE.
• Build benchmarks and regression tests to ensure performance improvements hold in production.
• Co-design kernels, runtimes, and models with hardware characteristics in mind.

Skills

CUDA kernels

Triton

Megatron

vLLM

TensorRT-LLM

NCCL

GPU performance

Distributed training

SGLang

Diffusion models

Education

Bachelor's degree in CS or related quantitative field

Tools

Nsight

CUTLASS

PTX

NVIDIA Nsight

NCCL

RDMA

InfiniBand

RoCE

Job description

Digitailor in Menlo Park, CA, is building high-performance AI/ML training and inference systems for image, video, and world-model workloads. You will optimize performance across GPU kernels, distributed execution, and multi-node serving.

The role emphasizes low-level GPU optimization with CUDA and Triton, designing distributed engines for diffusion models, and improving inter-GPU communication with NCCL, RDMA, InfiniBand, or RoCE.