Foundation
Models
Building models that understand, reason about, and interact with the physical world.
MODELS & LEARNINGR&D Director at X-Era Lab ↗

We released PhyAgentOS, a harness framework for self-evolving physical agents.
We joined LatePost for a talk about world action models.
01 / WHAT I WORK ON
Building models that understand, reason about, and interact with the physical world.
MODELS & LEARNINGBuilding the system layer that turns foundation models into reliable physical agents.
HARNESS & INFRASTRUCTUREBringing physical agents into real-world scenarios, from prototypes to scalable deployment.
APPLICATIONS & DEPLOYMENTSELECTED WORK · AGENT SYSTEMS
A self-evolving operating system
for physical agents.
02 / EXPERIENCE

2019–2022
Data-efficient learning for large-scale visual perception systems.

03 / BACKGROUND
Ph.D. at HCP Lab, supervised by Prof. Liang Lin. 2020–2025
B.Eng. in Computer Science and Technology. 2016–2020
TIP · 2024
IJCV · 2024
TPAMI · 2022
Conference reviewer
CVPR, ICCV, ICLR, NeurIPS, ICML, IJCAI, ACM MM, AAAI
Journal reviewer
TPAMI, TNNLS, TKDE, TOMM, TCSVT