U.S. Technical Lead at Wellwit Robotics

Robotics systems in the real world.

I work across mobile robotics, robot learning, and industrial automation, with a focus on AMR/AGV deployment, customer-site integration, debugging, and systems that hold up in real facilities.

Currently at Wellwit Robotics, I focus on North America technical execution for warehouse automation. I received my M.S. in Computer Science from Georgia Tech in May 2026, with robotics research experience in robot learning and dexterous manipulation.

Email CV GitHub LinkedIn
Linhao Bai in Atlanta
Atlanta / North America robotics work
Dexterous in-hand rotation
Video2Sim2Real pipeline
Video2Sim2Real pipeline
K-UBM pipeline
K-UBM behavior model
Working Stack

Where I spend my time

Customer-site robotics

Translate site workflows, operational constraints, and issue reports into robot behavior changes, deployment plans, and technical actions.

AMR / AGV systems

Mobile robot integration, warehouse automation, commissioning, debugging, and support across real facilities.

Robot learning bridge

Research experience in dexterous manipulation, visuomotor behavior models, simulation, and sim-to-real transfer.

North America execution

Technical ownership mindset: fast diagnosis, clear evidence, reproducible logs, and communication across customers, engineering, and operations.

Project Dossier

Robotics projects with deployable instincts

Video2Sim2Real pipeline figure
One human video Sim-to-real

Video2Sim2Real: Full-Stack Autonomous Dexterous Skill Acquisition from a Single Human Video

Yunhai Han, Jianuo Qiu, Linhao Bai, Ziyu Xiao, Zihang Zeng, Yangcen Liu, Zhaodong Yang, Shalin Jain, Wenrui Ma, Jiaqi Fu, Yuqian Zheng, Manisha Natarajan, Muhammad Zubair Irshad, Kenneth Shaw, Matthew Gombolay, Zsolt Kira, Harish Ravichandar

Reconstructs a simulator-ready digital twin from a single human manipulation video, refines robot behavior in simulation, and transfers the skill to real hardware.

Website arXiv Code Slides
K-UBM pipeline figure
Koopman Visuo-motor dexterity

Going with the Flow: Koopman Behavioral Models as Implicit Planners for Visuo-Motor Dexterity

Yunhai Han, Linhao Bai, Ziyu Xiao, Zhaodong Yang, Yogita Choudhary, Krishna Jha, Chuizheng Kong, Shreyas Kousik, Harish Ravichandar

Models dexterous skills as coupled visual and action flows, giving the robot a structured way to roll out temporally coherent behavior.

Website arXiv Video
Manipulation Control

In-Hand Rotation for Dexterous Manipulation

Linhao Bai

Studies stable object reorientation with multi-fingered hands under contact dynamics, sensing uncertainty, and hardware limitations.

Video Preview
Motion Lab

Robots should move before they persuade.

Local clip

In-hand rotation

Video2Sim2Real

Human video to robot

Video2Sim2Real

Real-world task clip

K-UBM

Policy comparison

K-UBM door replanning demonstration
K-UBM

Replanning under disturbance

Operations Log

The thread through the work

U.S. Technical Lead, Wellwit Robotics

Mobile robotics, robot learning, industrial automation, warehouse AMR/AGV deployment, customer-site integration, and North America technical execution.

Georgia Tech MSCS and STAR Lab

Robot learning, dexterous manipulation, visuomotor behavior models, simulation, and real-time robotic execution.

Robotics solutions and regional ownership

Build toward technical ownership for North America projects: evidence-driven support, deployment discipline, and systems that customers can trust.