Galaxy Yin (He Yin) | Amazon Robotics

I am Galaxy Yin, a robotics researcher and applied scientist working at the intersection of deep learning, whole-body control, and end-to-end navigation for humanoid robots.

My current interests include:

• Vision-Language-Action (VLA) models for humanoid robot navigation
• Video-Action-Models (VAM) for humanoid robot navigation
• Vision based whole body control for autonomous tasks
• RL based loco-manipulation policies for humanoid robot

I develop scalable loco-manipulation policies for humanoid robots, with hands-on experience solving sim-to-real challenges, and building teacher-student policy pipelines.

Recently, I have focused on VLA models for humanoid robot navigation. I have built the pipeline for collecting and annotating expert demonstration data, and improved model performance through supervised fine-tuning, outperforming both the expert policy and the baseline VLA. Also actively explored video-action-model-based approaches for navigation.

My background includes a PhD in Controls from UC Berkeley and work spanning RL, VLA, SLAM, and safety/stability analysis for complex robotic systems.

You can find my publications on my Google Scholar profile.