Robotics Research Scientist

Thespian Labs; Somerville, Northern (MA, KY); Hybrid; USD 120,000 - 190,000; Full time

Benefits offered by this job

Visa sponsorship

Equal opportunity employer

Job summary

Thespian Labs in Somerville, MA is seeking a Robotics Research Scientist to deploy our foundation model onto real robots. You will translate forward passes into perception, motion, and action on hardware in real time, ensuring safe, robust behavior.

You will connect perception, state estimation, and whole-body control with real hardware, running experiments on legible hardware, closing the sim-to-real gap, and iterating quickly with a small, hands-on team.

Qualifications

• PhD in robotics, machine learning, controls, or related field.

Responsibilities

• Bring the foundation model onto real robots to turn forward passes into perception, motion, and action on hardware in real time.
• Develop the control and whole-body motion that balance, reach, and gesture with learned behavior.
• Build perception and state estimation from real sensors to understand the room and the robot itself.
• Close the sim-to-real gap with simulation, domain randomization, and real hardware calibration.

Job description

Put a foundation model for embodied behavior onto real robots. Where the model meets the physical world, in real time.

Thespian Labs is an embodied intelligence research lab with roots at MIT. We are building a foundation model for behavior, the layer between reasoning and execution where a body has to do something coherent while the world keeps moving. Reasoning can plan and decide. Execution can render pixels and drive motors. The layer in between is the one nobody has built, and it is the whole of our work.

About the Role

As a Robotics Research Scientist at Thespian Labs, you will bring our model into the physical world. The behavior the model generates has to hold up on real hardware, where a body perceives a room, keeps its balance, reaches and walks, and acts while the world moves and does not wait. You will research and build the systems that take a single forward pass and turn it into a robot that moves with intent, safely and in real time.

You will work where learning meets control, connecting the foundation model to perception, state estimation, and whole-body motion on real robots, and closing the loop from a model on a screen to a humanoid acting in a room. The team is small and the loop is short, so you will run experiments on real hardware, find where physics breaks our assumptions, and feed what you learn straight back into the model. What you build has to work in the real world, on real hardware.

Responsibilities

• Bring the foundation model onto real robots, turning a single forward pass into perception, motion, and action on physical hardware.
• Develop the control and whole-body motion that let a robot balance, walk, reach, and gesture from learned behavior, in real time.
• Build the perception and state estimation a robot needs to make sense of a room, and of itself within it, from real sensors.
• Close the sim-to-real gap with simulation, domain randomization, and the calibration that makes a model trained off the robot work on it.
• Train robot behavior with reinforcement and imitation learning, and with data from teleoperation and real-world capture.
• Make behavior safe and robust on hardware, handling the edge cases, failures, and latency of acting in the real world.
• Run experiments on real robots and feed what physics teaches you back into the model, working alongside the research and engineering teams.

What We’re Looking For

• A PhD in robotics, machine learning, controls, or a related field, or an equivalent research record, with first-author work at venues like CoRL, RSS, ICRA, IROS, NeurIPS, ICML, or ICLR.
• Strong foundations in robot learning, classical robotics, or both, including control, motion planning, and state estimation, and how they meet modern deep learning.
• Strong programming skills in Python and fluency with ML frameworks like PyTorch, plus the engineering to make research run on real hardware.
• Hands‑on experience with real robots, simulation, or both, and a feel for where the two diverge.
• Demonstrated depth in one or more of reinforcement learning, imitation learning, whole‑body control, locomotion, manipulation, or robot perception.
• A bias for action and the appetite to drive hard, open‑ended problems on a small team, close to the hardware.

Preferred Qualifications

• Experience with humanoid robots or other legged or high‑degree‑of‑freedom platforms.
• Deploying learned policies on real hardware, not just in simulation.
• Familiarity with vision‑language‑action models or other approaches that connect large models to control.
• Teleoperation, data collection, or learning from demonstration on real robots.
• A background in real‑time systems, embedded software, or on‑device inference under tight latency.
• Substantial open‑source contributions, or research that made it onto a physical robot.

Thespian Labs is an equal‑opportunity employer. We hire for the work, and we want a lab full of people who don't all think alike. We sponsor visas for the right person.