Danfei Xu
Danfei Xu is a robotics researcher whose work helps robots learn from demonstrations, plan with language models, and transfer skills across robot types. He earned computer science degrees at Columbia and Stanford, then coauthored Neural Task Programming, DenseFusion, ProgPrompt, and Open X-Embodiment. In 2022 he joined Georgia Tech as an assistant professor while working as a research scientist at NVIDIA AI Research, and the National Science Foundation gave him a CAREER award in 2025. He is an assistant professor at Georgia Tech, a research lead at NVIDIA GEAR, and directs the Robot Learning and Reasoning Lab.
Timeline
-
2015–2021
Columbia and Stanford degrees
He earns a B.S. in computer science from Columbia University in 2015 and a Ph.D. in computer science from Stanford University in 2021.
-
2018
Neural Task Programming
In 2018, he coauthors Neural Task Programming (NTP), a framework that turns task specifications such as video demonstrations into hierarchical robot programs.
-
2019
DenseFusion object poses
In 2019, he coauthors DenseFusion, a framework that estimates the 6D poses of known objects from RGB-D images.
-
2022
Georgia Tech and NVIDIA
In 2022, he joins Georgia Tech’s School of Interactive Computing as an assistant professor and works as a research scientist with NVIDIA AI Research.
-
2022
ProgPrompt for robot plans
In 2022, he coauthors ProgPrompt, which uses program-like prompts to generate robot task plans for specific environments, robot capabilities, and tasks.
-
2023
Open X-Embodiment dataset
In 2023, he coauthors Open X-Embodiment, a robot dataset built from 22 robots across 21 institutions, and an RT-X model that transfers experience between robots.
-
2025
An NSF CAREER award
In 2025, the National Science Foundation gives him a CAREER award for research on models that give robots on-the-job training.
-
As of October 2026
Assistant professor and NVIDIA GEAR lead
As of October 2026, he is an assistant professor at Georgia Tech, a research lead at NVIDIA GEAR, and director of the Robot Learning and Reasoning Lab. His work connects robotics and machine learning. In 2026, he coauthors Human2Any, which learns reusable object-centric interaction priors from human videos for robot manipulation.
Career
- Since 2022Georgia Institute of TechnologyAssistant Professor
- Since 2022NVIDIAResearch Scientist
- NowRobot Learning and Reasoning LabDirector
Education
- 2021Stanford UniversityPh.D.
- 2015Columbia UniversityB.S.
Sources and credits
Photo: We have no published photo of Danfei Xu yet. Send us a free-licensed photo.
- Danfei Xu | School of Interactive Computing
- Neural Task Programming: Learning to Generalize Across Hierarchical Tasks
- DenseFusion: 6D Object Pose Estimation by Iterative Dense Fusion
- New Faculty Q&A with Danfei Xu | College of Computing
- ProgPrompt: Generating Situated Robot Task Plans using Large Language Models
- Open X-Embodiment: Robotic Learning Datasets and RT-X Models
- Two IC Faculty Receive NSF CAREER for Robotics and AR/VR Initiatives
- Danfei Xu
- Human2Any: Human-to-Robot Transfer via Constraint-Aware Compositional Planning



