AI undergraduate at Chongqing University
Exploring embodied intelligence, multi-agent world models, and collaborative decision-making under unreliable communication.
Contentment brings lasting happiness.
I am currently reproducing CoDreamer, a communication-based decentralised world model built on DreamerV3. My immediate goal is to establish a credible and auditable baseline before introducing controlled communication degradation and analysing where multi-agent coordination fails.
Research interests
- Embodied intelligence and cooperative agents
- Multi-agent reinforcement learning
- World models and model-based decision-making
- Communication and teammate modelling under uncertainty
| Project | What it is |
|---|---|
| My-CoDreamer | A from-scratch reproduction of CoDreamer based on the paper, with experiment records and reproducibility evidence. This is my current research-engineering focus. |
| codex-mac-calendar | A Swift plugin that connects Codex with the native macOS Calendar and Reminders apps. |
| geotrave | A multi-node collaborative travel-planning agent built with LangGraph, including the development history and lessons learned. |
| Code-Standards-Skills | Reusable AI instruction guides for more consistent work in specialised engineering tasks. |
- Worked on VR teleoperation in a university laboratory environment.
- Led digital-twin development in a Robocon team that received a national third prize.
- Completed an internship related to an Agent platform at Alibaba.
These experiences are being consolidated into reproducible projects, documented experiments, and clear accounts of my own contribution.
Languages: Python · TypeScript · Swift · Go
Engineering: Git · Docker · LangGraph
Currently learning through research: world models · model-based RL · multi-agent learning
If you are interested in multi-agent world models, embodied intelligence, or reproducible research engineering, feel free to get in touch.