Research Member of Technical Staff - Training Systems

About the Company

Rhoda AI is building the next generation of generalist intelligent robots. We own the full robotics stack from high-performance hardware and robot systems to the infrastructure and state-of-the-art foundation world models that control our robots. We've raised over $450M and are investing aggressively in model research, infrastructure, hardware development, and manufacturing scale-up.

Responsibilities

• Own training performance end-to-end and diagnose performance of large-scale multimodal training
• Build systematic performance attribution including step-time decomposition and scaling curves
• Drive gains in distributed efficiency, compute efficiency, and memory efficiency
• Design and evolve parallelism strategies including data, tensor, pipeline, and sharding approaches
• Establish source-of-truth performance metrics such as MFU, throughput, and scaling efficiency
• Build tools for bottleneck identification, performance tracking, and regression detection
• Partner with research scientists to translate model innovations into scalable implementations
• Collaborate with infrastructure teams to improve cluster-level utilization and network performance

Requirements

• Proven track record improving large-scale distributed training performance
• Deep hands-on experience with PyTorch
• Strong understanding of data, tensor, and pipeline parallelism
• Experience with sharded training (FSDP / ZeRO-style) and communication patterns
• Strong systems intuition regarding compute, communication, and memory bottlenecks
• Exceptional debugging and measurement abilities
• High ownership mindset in fast-moving environments

Preferred Qualifications

• Experience with JAX
• GPU kernel or compiler-level experience (CUDA, Triton, graph capture, operator fusion)
• Experience with multimodal or video training
• Experience with large-scale training frameworks or distributed runtimes
• Familiarity with cluster topology, networking, and large-scale scheduling