OpenAI Marcin Andrychowicz

About

Marcin Andrychowicz is a leading researcher in artificial intelligence, best known for his groundbreaking work in reinforcement learning (RL) and dexterous robotic manipulation. His most impactful contribution is the 2019 paper "Learning dexterous in-hand manipulation," which has amassed over 1,588 citations. In this seminal work, Andrychowicz and his team at OpenAI demonstrated how RL could train a physical Shadow Dexterous Hand to perform complex, vision-based object reorientation tasks entirely in simulation. By employing extensive domain randomization—varying physical properties like friction, mass, and lighting—the learned policies transferred zero-shot to the real world, achieving unprecedented dexterity. This paper not only showcased the power of sim-to-real transfer but also set a new benchmark for robotic manipulation. Beyond this, Andrychowicz has contributed to foundational RL algorithms, including Hindsight Experience Replay, which accelerates learning in sparse-reward environments. His work bridges the gap between simulated training and real-world deployment, inspiring a generation of researchers to tackle high-dimensional control problems. For students and researchers, Andrychowicz’s research exemplifies how creative problem-solving and robust simulation can unlock capabilities once thought impossible for robots.

Research Focus

Key Achievements

Top Papers

1. 1
   Learning dexterous in-hand manipulation

   1,588 citations · 2019

Key Collaborators

Contact & Links