Ondrej Biza
I am a researcher at the Boston Dynamics AI Institute in Cambridge, MA.
My research goal is to build generalist robots for household and industrial settings. I place my research in the areas of (geometric) deep learning, reinforcement learning, robot learning and robotic manipulation. Most recently, I have been working on bimanual policy learning. I am especially interested in the confluence of data, from large internet datasets, from teleoperation, and from on-robot execution. How do we combine these datasets to train bimanual policies that generalize across object and environments, and retain high precision?
I was fortunate to intern at Google Brain in 2022 and the Boston Dynamics AI Institute in 2023-2024. I recevied my PhD in Computer Science from Northeastern University in 2024 and my undergraduate degree from Czech Technical University in 2019.
at gmail dot com
news
- Oct, 2024 I joined the Boston Dynamics AI Institute as a researcher.
- Sep, 2024 I defended my PhD thesis!
- Sep, 2023 I presented One-shot Imitation Learning via Interaction Warping at CoRL'23 in Atlanta.
- Jul, 2023 I presented Invariant Slot Attention at ICML'23 in Honolulu.
selected publications
On-Robot Reinforcement Learning with Goal-Contrastive Rewards
Ondrej Biza, Thomas Weng, Lingfeng Sun, Karl Schmeckpeper, Tarik Kelestemur, Yecheng Jason Ma, Robert Platt, Jan-Willem van de Meent, Lawson L. S. Wong. Preprint, 2024.
