Remote Deep Learning Engineer - Distributed Training Scale

BairesDev; Peru (IL); On-site; USD 120,000 - 180,000; Full time

Benefits offered by this job

Remote work

Competitive USD compensation

Home setup provided

Flexible hours

Parental leave & holidays

Mentorship & growth opportunities

Job summary

BairesDev invites an experienced Deep Learning Engineer (Distributed) to lead large-scale multi-GPU training. You will design distributed pipelines, optimize throughput, and ensure scalable checkpoints across compute clusters.

You will work with PyTorch, DDP/FSDP/DeepSpeed, Triton/TensorRT, and profiling tools like Nsight, collaborating with research and systems teams to push performance and reliability at scale.

Qualifications

• 4+ years of experience in ML engineering, distributed systems, or related field.
• Proven PyTorch experience with DDP, FSDP, and DeepSpeed.
• Strong CUDA expertise and Nsight profiling.
• Experience with Triton or TensorRT for performance optimization.
• Advanced proficiency in English.

Responsibilities

• Design and implement distributed training pipelines using PyTorch DDP, FSDP, and DeepSpeed.
• Optimize model performance through mixed-precision training and advanced sharding techniques.
• Identify and resolve system bottlenecks using Nvidia GPU profiling tools like Nsight.
• Develop and deploy optimized model kernels using Triton or TensorRT to enhance inference and training speed.
• Collaborate with research and systems teams to manage distributed state and resilient checkpointing at scale.

Skills

PyTorch

Distributed training frameworks (DDP,F

Tools

Nsight

Triton

TensorRT

Job description

BairesDev invites an experienced Deep Learning Engineer (Distributed) to lead large-scale multi-GPU training. You will design distributed pipelines, optimize throughput, and ensure scalable checkpoints across compute clusters.

You will work with PyTorch, DDP/FSDP/DeepSpeed, Triton/TensorRT, and profiling tools like Nsight, collaborating with research and systems teams to push performance and reliability at scale.