Georgy Savva

georgy.savva@nyu.edu New York, NY, United States

I am a first-year Master’s student in Computer Science at NYU Courant Institute of Mathematical Sciences. I study machine learning. My research interests lie in multimodal learning, representation learning, image generation, and reinforcement learning.

Previously, I interned with Prof. Lerrel Pinto working on robot learning at the NYU CILVR lab. Before that, I spent eight years in industry as a software engineer, specializing in backend engineering and distributed systems. I was a founding engineer at Edwin.ai, a Y Combinator-backed EdTech startup. Additionally, I created an open-source library Scany and served as the technical co-founder of Scifind, a troubleshooting platform for bioscientists. I received my Bachelor’s in Computer Science from the Russian University of Cooperation.

selected publications

HuDOR: Bridging the Human to Robot Dexterity Gap through Object-Oriented Rewards

Irmak Guzey, Yinlong Dai, Georgy Savva, and 2 more authors

(Under Review), 2024

Abs PDF Website

Training robots directly from human videos is an emerging area in robotics and computer vision. While there has been notable progress with two-fingered grippers, learning autonomous tasks without teleoperation remains a difficult problem for multi-fingered robot hands. A key reason for this difficulty is that a policy trained on human hands may not directly transfer to a robot hand with a different morphology. In this work, we present HUDOR, a technique that enables online fine-tuning of the policy by constructing a reward function from the human video. Importantly, this reward function is built using object-oriented rewards derived from off-the-shelf point trackers, which allows for meaningful learning signals even when the robot hand is in the visual observation, while the human hand is used to construct the reward. Given a single video of human solving a task, such as gently opening a music box, HUDOR allows our four- fingered Allegro hand to learn this task with just an hour of online interaction. Our experiments across four tasks, show that HUDOR outperforms alternatives with an average of 4× improvement. Code and videos are available on our website https://object-rewards.github.io/.