Generalist
Research Scientist: Pretraining
$240K–$350K
San Francisco Bay Area (San Mateo) · On-site · Full-time · Posted Feb 12, 2026
- Experience
- Senior/staff (L6+) experience with large-scale ML systems; deep experience training large models at scale.
About the job
Design and run large-scale pretraining for multimodal robot foundation models that learn general capabilities across embodiments, tasks, and environments.
What you'll do
- Design and execute large-scale pretraining runs; define model architectures, objectives, and training curricula.
- Develop data mixtures and sampling strategies for petabyte-scale datasets, and guide data collection and dataset sourcing.
- Run ablations on scaling, data quality, and architecture; collaborate with ML Infra and Systems to improve training performance and reliability.
What they ask for
- Deep experience training large transformer or diffusion models at scale.
- Led or significantly contributed to multi-node, multi-GPU distributed training.
- Experience with scaling laws, optimization dynamics, and large-model failure modes.
- Strong PyTorch fundamentals and ability to debug across the stack.
Pay
$240K – $350K USD base salary
Tools
transformer, diffusion, PyTorch, ML Infra, Systems
AI focus
Pretraining multimodal robot foundation models using video, proprioception, action traces, language, and other robotic data to develop general-purpose capabilities.
Model-written summary · the original posting is authoritative