Research Scientist, Robotics

Job description

**About Tutor** We believe robots are the biggest lever we have to making the physical world a better, more affordable place. But robots are a rounding error in our society: nearly nine out of ten American factories and warehouses don’t even have a single robot, and robots have yet to make their way into our homes or places of business. Tutor Intelligence is changing that. Tutor is a robotics research and deployment company, building generally capable robot hardware and artificial intelligence to enable robots to go where no robots have gone before. [Sonny](https://tutorintelligence.com/sonny) is a semi-humanoid robot with frontier dexterity approaching the human form. [Cassie](https://tutorintelligence.com/cassie) does back-breaking work, moving boxes, pallets, and bulk goods for businesses from family-owned manufacturers to some of the largest brands in the world. Tutor robots are manufactured at scale and operate as a network, combining the whole fleet’s learned experience to train ever more capable robot artificial intelligence. Founded at MIT and headquartered in Watertown, MA, our team is full stack: we design and manufacture robot hardware, train our own models, work directly with the smallest and largest end-customers, and maintain our fleet across the United States. We’re a team of researchers, engineers, technicians, operators, and sellers who love building things and solving hard problems. Our mission is grand: robots everywhere, for the common good, raising the standard of living across North America, and then the world. **The Role** Tutor's research team pushes the frontier of robot capability across the full stack. Researchers have access to [100+ robots](https://tutorintelligence.com/blog/building-a-100-robot-data-factory-toward-factory-ready-ai) for data collection and evaluation, a massive network of teleoperators, scaled hardware and manufacturing infrastructure, and the ability to collect real world data at enormous scale. Our work covers the complete lifecycle of real-world data generation and model training from initial behavior cloning to post-training with supervised fine-tuning and reinforcement learning. And researchers and research engineers work closely with applied engineers to develop novel techniques into robot products that can be deployed at scale. As a research scientist, you'll set the direction for how our robots learn. ### What you'll work on - Set research directions in robot learning for dexterous manipulation, from framing the problem to results that hold up on robots. - Develop methods that learn from large-scale human supervision: teleoperated demonstrations, interventions and corrections, and human feedback used as reward. - Design experiments across the DF1 fleet, and build the evaluations that separate a real improvement from noise. - Train and post-train vision-language-action models with behavior cloning, supervised fine-tuning, and reinforcement learning. - Work with research and applied engineers to turn new techniques into models that run on deployed Cassie and Sonny robots. - Share results through internal write-ups and technical reports. ### Requirements - 5+ years of research experience in robot learning, machine learning, or a related field, through a PhD or equivalent industry experience. - A track record of original work in robot learning, imitation learning, reinforcement learning, or large-model training, shown through publications, open-source work, or shipped systems. - Strong Python and hands-on experience with PyTorch or JAX. - Experience running experiments on physical robots, not only in simulation. - Clear communication of results, including the negative ones. ### Nice to haves - Vision-language-action models, or diffusion and flow-matching policies. - Learning from human feedback, interventions, or reward modeling. - Bimanual or dexterous manipulation. - Large-scale distributed training. - Not afraid to step foot in a factory or a warehouse.