Robotics and AI Institute logo

Sim2Real Robotics Research Scientist

Robotics and AI Institute

Cambridge, MAOnsitePosted 1 day ago$140,910 - $326,040/year
Full TimeSenior LevelResearch & ScientistResearch Institute
Original posting ↗

Summary

The RAI Dexterous Manipulation team is developing advanced manipulation systems capable of performing precise, contact-rich tasks at or above human cadence. By co-designing hardware, software, controls, and machine learning components in an integrated fashion, we seek to push the state of the art to enable reliable, high-performance skills across complex manufacturing and construction tasks capable of real-world deployment.

Job description

Mission

The RAI Dexterous Manipulation team is developing advanced manipulation systems capable of performing precise, contact-rich tasks at or above human cadence. By co-designing hardware, software, controls, and machine learning components in an integrated fashion, we seek to push the state of the art to enable reliable, high-performance skills across complex manufacturing and construction tasks capable of real-world deployment.

Responsibilities

  • Develop, deploy, test, and debug reinforcement learning (RL) policies on real-world novel robotic hardware platforms for dexterous manipulation tasks.
  • Investigate, analyze, and mitigate domain gaps between physics simulation and real-world execution for contact-rich manipulation through parameter identification, dynamic domain randomization, and real2sim alignment to continuously improve simulator fidelity.
  • Design and implement curriculum training frameworks, demonstration-led state guidance, and reward engineering strategies to overcome exploration bottlenecks in tight-tolerance manipulation tasks.
  • Integrate multi-modal sensor inputs—including vision, tactile/contact sensing, 6-axis force/torque, and joint proprioception—into real-time sensorimotor learning architectures and policies.
  • Explore policy learning techniques for highly-dynamic, contact-rich action spaces capable of regulating speeds, impedances, and energy expenditure.
  • Develop high-throughput experimental workflows and collaborate closely with research scientists, hardware engineers, and control teams to co-design closed-loop learning systems and maintain technical documentation.

Requirements

  • M.S. or Ph.D. in Robotics, Computer Science, Machine Learning, or a related field.
  • Hands-on expertise with real-world robot integration, hardware deployment, physical testing setups.
  • Experience in developing, adapting, fine-tuning, or deploying multi-modal policies integrating vision and contact-rich physical modalities (tactile or force/torque sensing).
  • Solid background in deep reinforcement learning (RL) algorithms, modern simulation-to-real (Sim2Real) transfer methodologies, reward engineering/shaping, and curriculum design.
  • Experience in incorporating state-of-the-art workflows and techniques to automate robotics training loops, simulation environments, or Sim2Real transfer tuning.
  • Demonstrated experience in Sim2Real techniques for either manipulation or locomotion is required, with explicit investigation and application of these methods to dexterous, contact-rich robotic manipulation considered a preference.
  • Proficiency in Python, deep learning frameworks, and physics simulation engines (e.g., Physx, MuJoCo, Netwon) applied to complex robotic systems.

Requirements

  • M.S. or Ph.D. in Robotics, Computer Science, Machine Learning, or a related field
  • Hands-on expertise with real-world robot integration, hardware deployment, physical testing setups
  • Experience in developing, adapting, fine-tuning, or deploying multi-modal policies integrating vision and contact-rich physical modalities (tactile or force/torque sensing)
  • Solid background in deep reinforcement learning (RL) algorithms, modern simulation-to-real (Sim2Real) transfer methodologies, reward engineering/shaping, and curriculum design
  • Experience in incorporating state-of-the-art workflows and techniques to automate robotics training loops, simulation environments, or Sim2Real transfer tuning
  • Demonstrated experience in Sim2Real techniques for either manipulation or locomotion
  • Proficiency in Python, deep learning frameworks, and physics simulation engines (e.g., Physx, MuJoCo, Netwon) applied to complex robotic systems

Benefits

  • Annual Bonus
  • Long-Term Cash Incentive
  • Cell Phone Reimbursement
  • Commuter Subsidy
  • Medical Benefits

Company overview

The Robotics and AI Institute (RAI Institute) is a cutting-edge research organization dedicated to solving fundamental problems in artificial intelligence and robotics. Headquartered in Cambridge, Massachusetts with a Swiss subsidiary in Zurich, the institute brings together top talent across robotics, perception, AI, machine learning, computer science, and engineering to develop future generations of intelligent machines. Founded by Marc Raibert, the visionary behind Boston Dynamics, the RAI Institute operates at the intersection of academic research and industrial development, combining the best features of university research labs with corporate development environments. The institute focuses on four core research areas: cognitive AI, athletic AI, organic hardware design, and robotics ethics. Through strategic partnerships with industry leaders like Boston Dynamics, the RAI Institute advances humanoid robotics through reinforcement learning and develops technology aimed at increasing productivity, enhancing human safety, supporting people with disabilities, and improving quality of life. The organization's work represents the forefront of intelligent machine development, pushing the boundaries of what robots can achieve through innovative research and collaborative development.

Explore more jobs of your matchFind more biopharma roles with GetBioJob.