Research Scientist (San Francisco)

Research Scientist, Post-Training | Robotics & AI | SF Bay Area or Boston A frontier AI robotics research and product company is building general-purpose robots. Their tech can work alongside humans in industrial and home environments. Their team comes from OpenAI, Boston Dynamics, Google DeepMind , and other frontier labs, with hands-on experience shipping some of the most recognised AI and robotics products of the last decade.

They're working at the intersection of large-scale AI and physical robotics, with a focus on dexterity and embodied foundation models.

They're hiring a post-training research scientist to turn large pretrained robot models into systems that actually work in the real world.

The work

• Pretraining gives you a capable model. Post-training makes it reliable, controllable, and deployable. This role owns that second half - fine-tuning and adapting models for specific robotic tasks and embodiments, developing methods for robustness and controllability, and building evaluation frameworks that measure real-world performance rather than just offline metrics.
• You'll also work on inference-time performance (latency, stability, memory footprint) and close the loop between model outputs and physical outcomes using techniques like imitation learning, RL, distillation, synthetic data, and curriculum learning.
• Over time, you'll grow into a full-stack ML roboticist - comfortable debugging from loss curves all the way through to robot behaviour.

What they're looking for

Someone with experience fine-tuning large models for downstream tasks (RLHF, IL, RL, distillation, domain adaptation), ideally with exposure to embodied AI or real-world ML systems. You care about evaluation and failure analysis, enjoy rapid iteration with physical feedback loops, and want to bridge the gap between foundation models and deployment on actual hardware.

On-site - SF Bay Area or Boston

Up to $350,000 + equity