Lead ML Training Systems Engineer - Multimodal, Large-Scale

Rhoda AI; Palo Alto (CA); On-site; USD 210,000 - 320,000; Full time

Job summary

Rhoda AI is building next-generation generalist robots and is seeking a Staff / Principal ML Training Systems Engineer to own training systems performance end-to-end. You will optimize large-scale multimodal training, define parallelism strategies, and drive efficiency across GPUs, memory, and compute.

This is a core, research-facing systems role with high impact on model iteration velocity. You will work with researchers to translate innovations into scalable implementations, improving

Qualifications

• Experience improving large-scale distributed training performance across multimodal data.
• Hands-on expertise with modern ML stacks (PyTorch required; JAX a plus).
• Strong understanding of data/tensor/pipeline parallelism, sharded training, and scaling across GPUs.

Responsibilities

• Own training performance end-to-end for large-scale multimodal models.
• Diagnose bottlenecks and drive measurable efficiency gains in compute, communication and memory.
• Design and evolve parallelism strategies and training frameworks to improve speed and scalability.
• Collaborate with researchers to translate model innovations into scalable, efficient implementations.

Skills

Large-scale distributed training

PyTorch

JAX

Systems optimization

Performance profiling

Tools

CUDA

Triton

Job description

Rhoda AI is building next-generation generalist robots and is seeking a Staff / Principal ML Training Systems Engineer to own training systems performance end-to-end. You will optimize large-scale multimodal training, define parallelism strategies, and drive efficiency across GPUs, memory, and compute.

This is a core, research-facing systems role with high impact on model iteration velocity. You will work with researchers to translate innovations into scalable implementations, improving