- Co-founder and CEO, Skild AI1
- Raj Reddy Associate Professor, School of Computer Science (Robotics Institute; affiliated with the Machine Learning Department), Carnegie Mellon University1
Deepak Pathak is a co-founder and the CEO of Skild AI, which is developing a foundation model for robotics, and the Raj Reddy Associate Professor in Carnegie Mellon University's School of Computer Science. Pathak's research sits at the intersection of computer vision, machine learning and robotics, with a focus on agents that learn from raw sensory data.1
At a glance
- Current roles
- Co-founder and CEO +1 more
- Companies
- Skild AI, Carnegie Mellon University
- Sector
- Company
- Education
- Carnegie Mellon, UC Berkeley
- Work listed
- 2
- Honors
- 5
- Sources cited
- 6
- Profile reviewed
Career
Pathak's site lists a PhD from UC Berkeley advised by Alyosha Efros and Trevor Darrell, a visiting postdoctoral appointment at UC Berkeley with Pieter Abbeel, and a year as a researcher at Meta AI Research. Pathak now holds the Raj Reddy chair at Carnegie Mellon, awarded in 2024, as a member of the Robotics Institute.1
As co-founder and CEO of Skild AI, Pathak leads work on a single robot "brain" meant to operate many robot bodies and tasks. Skild AI has introduced S1, a robotic foundation model built as an in-context learner that executes a task after being shown a video.12
Research
Pathak's group draws on psychology to build systems that learn using data as their own supervision. Early work on curiosity-driven exploration used prediction error as an intrinsic reward, and later work with students covered legged locomotion over difficult terrain using egocentric vision.134
Skild AI's technical writing on its general-purpose robot brain cites several of Pathak's papers, including work on zero-shot visual imitation and on learning robot affordances from human videos.5
Work and contributions
Curiosity-driven exploration by self-supervised prediction
First author of a 2017 paper that uses an agent's error in predicting the consequences of its own actions as an intrinsic reward for exploration.3
LocoFormer
Co-author of a 2025 paper on a generalist locomotion policy that adapts by accumulating experience in a long context, which Skild AI cites as its in-context learner for locomotion.62
Timeline
- First-authored the paper on curiosity-driven exploration by self-supervised prediction.3
- Co-authored a paper on legged locomotion in challenging terrains using egocentric vision.4
- Was awarded the Raj Reddy Chair at Carnegie Mellon University.1
- Skild AI published its account of building a general-purpose robotic brain.5
- Co-authored the LocoFormer paper on generalist locomotion via long-context adaptation.6
Other documented roles
Roles the sources document that are not confirmed as current.
- Researcher for one year, collaborating with Jitendra Malik, Meta AI Research1
Honors and recognition
Where they work
- Skild AI 1 person in the wiki
- Carnegie Mellon University 24 people in the wiki
Related people
- Justin FuReinforcement learning · Robotics
- Pieter AbbeelRobotics · Reinforcement learning
- Jonathan HurstRobotics · Reinforcement learning
- Amnon ShashuaComputer vision · Robotics
- Yann LeCunComputer vision · Robotics
- Chelsea FinnRobotics · Reinforcement learning
- Bob McGrewRobotics · Reinforcement learning
- Byron BootsRobotics · Computer vision
- Sergey LevineRobotics · Reinforcement learning
- Josef SivicComputer vision · Robotics
Suggest a correction or removal
Profiles cover public professional work only, and each statement cites a source. If something is out of date, missing or wrong, tell us and link a source. People can also ask for their profile to be corrected or removed.
References
6 sources · profile reviewed Oct 21, 2026 · data updated Oct 21, 2026
- Deepak Pathak — homepage (Carnegie Mellon University)
- Introducing S1: In-Context Learning for Robotics
- Curiosity-driven Exploration by Self-supervised Prediction (arXiv 1705.05363)
- Legged Locomotion in Challenging Terrains using Egocentric Vision (arXiv 2211.07638)
- Building the general-purpose robotic brain
- LocoFormer: Generalist Locomotion via Long-context Adaptation (arXiv 2509.23745)