Research Engineer - Distributed Training

What you'll do

• Improve end-to-end training efficiency across compute, memory, networking, and scheduling.
• Develop low-level performance optimizations and distributed training systems for data, tensor, and pipeline parallel workloads.
• Help shape RL training architecture, including asynchronous rollout and post-training systems, and contribute to open-source and internal infrastructure.

What they ask for

• Strong systems engineering experience in AI/ML infrastructure, especially large-scale model training or inference.
• Deep familiarity with PyTorch and distributed training frameworks or related tooling.
• Experience optimizing training performance across kernels, memory movement, communication overhead, or parallelization.
• Hands-on experience with data, tensor, and pipeline parallelism; understanding of GPU architecture, profiling, and performance debugging.

Nice to have

• Experience writing or optimizing CUDA / Triton kernels.
• Experience with compiler or runtime optimization, RL training infrastructure, multi-node GPU clusters, or open-source ML systems.

Pay

$150-350k (total pay)

Tools

PyTorch, PyTorch Distributed, DeepSpeed, FSDP, Megatron, vLLM, Ray, CUDA / Triton

AI focus

Build infrastructure for pre-training and large-scale RL training of frontier models, including asynchronous rollout and post-training systems.