Tianci Wang - Embodied AI & Agentic AI Researcher

About Me 👋🏻

👋🏻 I am a Master's student at the School of Artificial Intelligence and Data Science (AI&DS), University of Science and Technology of China (USTC) , working at the SCAI LAB and advised by Prof. Jingrun Chen and Prof. Congcong Zhu.

🔬 My research focuses on Large Language Models (LLMs), Embodied AI and Agentic AI, with a particular focus on agent-driven physical simulation.

🚀 I am passionate about building LLM-powered autonomous agents that plan, generate, and iteratively self-correct physics simulations, producing verifiable, ground-truth trajectory data to train and evaluate embodied AI systems.

📫 Feel free to reach out to me at wangtianci2004@gmail.com or wangtianci2004@mail.ustc.edu.cn.

News 📢

[2025.09] Admitted to the School of AI&DS, USTC (via postgraduate recommendation). 🎉 Honored to be advised by Prof. Jingrun Chen and Prof. Congcong Zhu! ✨
[2025.07] Two papers was accepted by ICML 2025 Workshops. ✨ See you in Vancouver, Canada!
[2025.07] One paper was accepted by IEEE Computer Graphics and Applications. 🎉
[2025.04] One paper was accepted by IEEE PacificVis 2025. 🎉 See you in Taipei, China!

Open Source Projects 🛠️

PhysPi — AI-Native Physics Simulation Agent

PhysPi turns natural-language scene descriptions into executable physics simulations, rendered videos, and verifiable trajectory data. It combines an LLM-powered agent layer with the Genesis physics engine to plan scenes, generate simulation code, validate runs, inspect artifacts, and iteratively self-correct toward reproducible results.

Code (Coming soon)

WeJH-Taro — ZJUT Campus Mini-Program

A Taro + Vue 3 frontend for WeJH, a unified campus mobile portal for Zhejiang University of Technology students. It brings together course schedules, grades, exams, available classrooms, campus-card services, and library features in one mini-program.

Code