SlipstreamJobs tracks this role from the company's public career site. Apply directly on the employer's site.
Generalist is building embodied foundation models for robotics, focusing on dexterity and enabling robots to intelligently interact with the physical world. The company combines large-scale AI and robotics, with a team drawn from OpenAI, Boston Dynamics, Google DeepMind, and other frontier labs.
As a Research Scientist in Post-Training, you will transform pretrained robot models into production-ready systems. Post-training is where research meets reality—taking general models and making them useful, controllable, safe, and performant in real-world robotic applications.
Key responsibilities include:
- Designing fine-tuning and adaptation strategies for downstream robotic tasks and different robot embodiments
- Developing methods to improve reliability, robustness, and controllability of deployed models
- Building evaluation frameworks that measure actual robot performance in the physical world, not just offline metrics
- Optimizing inference-time performance (latency, stability, memory footprint) in collaboration with ML infrastructure teams
- Applying techniques such as imitation learning, reinforcement learning, distillation, synthetic data generation, and curriculum learning
- Closing the feedback loop between model outputs and physical-world outcomes
Ideal candidates have experience fine-tuning large models for downstream tasks (RLHF, IL, RL, distillation, domain adaptation), background in embodied AI or robotics, and comfort debugging across the full ML stack. You should care deeply about evaluation and failure analysis, enjoy rapid iteration with real-world feedback, and want to bridge foundation models and physical deployment.
