Vikash Kumar

Career

Kumar completed a PhD in Computer Science and Engineering at the University of Washington, advised by Emanuel Todorov and Sergey Levine, with a thesis on manipulators and manipulation in high-dimensional spaces. Kumar's CV then lists a role as a member of technical staff at OpenAI in 2017, a research scientist position at Google Brain from 2018 to 2019, and a research scientist role at FAIR (Meta AI) from 2020 to 2023.²⁵

Since January 2023 Kumar has been an Adjunct Professor at the Carnegie Mellon University Robotics Institute, and since May 2023 the co-founder and CEO of MyoLab.ai. MyoLab describes its work as building human-aware embodied intelligence, including a foundation model it calls MyoSapiens.²¹⁶

Research

Kumar's research looks at embodied movement across biological, digital and electromechanical systems, combining machine learning, biomechanics, robotics and optimal control. The doctoral work produced the ADROIT manipulation platform, a 28-degree-of-freedom arm-and-hand system driven by custom pneumatic actuation, together with methods for synthesizing dexterous behavior on physical hardware.¹⁵

Kumar's publication list spans dexterous manipulation, model-based and multi-task robot learning, visual representations for robots such as R3M, and benchmarks such as ROBEL, a set of low-cost robots for learning research introduced in 2019 with colleagues including Sergey Levine.⁷⁸

Open-source work

Kumar is listed as lead creator of MyoSuite and RoboHive and as a founding member of the MuJoCo physics engine and the TorchRL library. MyoSuite, introduced in a 2022 paper, offers contact-rich musculoskeletal motor-control tasks simulated in MuJoCo, and RoboHive, described in a 2023 paper with Kumar as first author, gathers robot-learning environments under one framework.^19101134

Work and contributions

• MyoSuite Open-source code

  Listed by Kumar's homepage as lead creator; Kumar is an author of the MyoSuite paper (L4DC 2022) that the Apache-2.0 repository asks users to cite. MyoSuite provides musculoskeletal environments and tasks simulated with MuJoCo.¹³

• RoboHive Open-source code

  Listed by Kumar's homepage as lead creator of RoboHive, a unified framework for robot learning with MuJoCo-simulated environments exposed through the OpenAI Gym API, published under Apache-2.0 from Kumar's GitHub account.¹⁴

Timeline

1. 2016Received the ICRA 2016 Best Manipulation Paper award for work on optimal control with learned local models for dexterous manipulation.⁷²
2. Feb 2017Completed a PhD in Computer Science and Engineering at the University of Washington.²
3. Apr 2017Joined OpenAI as a member of technical staff.²
4. Feb 2018Joined Google Brain as a research scientist.²
5. Sep 2019Co-authored ROBEL, a set of robotics benchmarks for learning with low-cost robots.⁸
6. Nov 2020Joined FAIR at Meta AI as a research scientist.²
7. May 2022Released the MyoSuite paper on a contact-rich simulation suite for musculoskeletal motor control.¹⁰
8. Jan 2023Became an Adjunct Professor at the Carnegie Mellon University Robotics Institute.²
9. May 2023Co-founded MyoLab.ai and became its CEO.²
10. Oct 2023Published RoboHive, a unified framework for robot learning, as first author.¹¹
11. 2024Gave an Early Career Keynote at the Conference on Robot Learning (CoRL 2024).²¹

Other documented roles

Roles the sources document that are not confirmed as current.

• Research Scientist, FAIR–Meta AI (Nov 2020 – Apr 2023), Meta¹²
• Research Scientist, Google Brain (Feb 2018 – Aug 2019), Google²⁹
• Member of technical staff (Apr 2017 – Oct 2017), OpenAI²

Honors and recognition

• Best Manipulation Paper Award, ICRA 2016 · IEEE International Conference on Robotics and Automation, 2016²⁷
• Google Cloud Research Award · Google, 2019²
• Best Paper Award in Robot Manipulation finalist, ICRA 2024 · IEEE International Conference on Robotics and Automation, 2024²⁷

Where they work

• Myolab.ai 1 person in the wiki
• Carnegie Mellon University 24 people in the wiki

Related people

• Misha DenilRobotics · Reinforcement learning
• Justin FuReinforcement learning · Robotics
• Chelsea FinnRobotics · Reinforcement learning
• Joelle PineauReinforcement learning · Robotics
• Bob McGrewRobotics · Reinforcement learning
• Gavriel StateRobotics · Reinforcement learning
• Tuomas HaarnojaReinforcement learning · Robotics
• Sergey LevineRobotics · Reinforcement learning
• Jonathan HurstRobotics · Reinforcement learning
• Deepak PathakRobotics · Reinforcement learning