Generalist
Research Scientist: Post-Training
$240K–$350K
San Francisco Bay Area (San Mateo) · On-site · Full-time · Posted Sep 2, 2026
About the job
Train and adapt pretrained robot models into production-ready systems, validating their performance on physical robots.
What you'll do
- Design fine-tuning and adaptation strategies for robotic tasks and embodiments; improve reliability, robustness, and controllability.
- Build real-world evaluation frameworks and analyze failures; improve inference latency, stability, and memory footprint.
- Use methods including imitation learning, reinforcement learning, distillation, synthetic data, and curriculum learning, connecting model outputs to physical outcomes.
What they ask for
- Experience fine-tuning large models for downstream tasks, such as RLHF, IL, RL, distillation, or domain adaptation.
- Experience in embodied AI, robotics, or real-world ML systems.
Nice to have
- Care about evaluation, benchmarking, and failure analysis.
- Comfort debugging across the ML stack and iterating with real-world feedback.
Pay
$240K – $350K (base salary)
Tools
RLHF, IL, RL
AI focus
Post-train and adapt embodied foundation models for robotic tasks, including fine-tuning, reinforcement learning, human feedback, evaluation, and on-robot validation.
Model-written summary · the original posting is authoritative