Lizhe Chen 陈立哲
M.S. Student · Interactive Media Technology · Tsinghua SIGS 硕士研究生 · 互动媒体技术 · 清华大学深圳国际研究生院
Hey there! I'm a graphics nerd who fell into the deep end of AI. I spend my days hacking on Vulkan engines and GPU renderers, my nights training models to reason and see — and somewhere in between I ship indie games with friends. If it runs on a GPU and makes pixels or predictions, I'm probably into it. 你好呀！我是一个掉进 AI 深坑的图形学爱好者。白天写 Vulkan 引擎和 GPU 渲染器，晚上训大模型做推理和多模态理解，现在正在做一些独立游戏。只要跑在 GPU 上、能出画面或出结果的东西，我都感兴趣。
Real-time Rendering & Neural Textures 实时渲染与神经纹理
GPU-accelerated pipelines, PBR, real-time ray tracing (OptiX), neural texture synthesis, 3D Gaussian Splatting, and Vulkan engine development. GPU加速渲染管线、物理着色、实时光追（OptiX）、神经纹理合成、3DGS与Vulkan引擎开发。
LLM & VLM Reasoning 大语言模型与视觉大模型推理
Chain-of-thought reasoning, in-context learning, multimodal understanding, prompt compression, and graph reasoning with large language models. 思维链推理、上下文学习、多模态理解、提示压缩，以及基于大语言模型的图推理。
Education 教育经历
Tsinghua University, SIGS 清华大学 深圳国际研究生院
2025 — PresentM.S. in Interactive Media Technology 互动媒体技术方向 · 硕士研究生
Game engine R&D, 3D data synthesis, procedural generation, physically-based rendering. 游戏引擎研发、三维数据合成、程序化生成与基于物理渲染。
North China University of Technology 北方工业大学
2021 — 2025B.S. in Digital Media Technology · Rank 1st 数字媒体技术 · 工学学士 · 绩点排名第一
Work 工作经历
DexForce Intelligent Digital Tech 跨维（深圳）智能数字科技
Mar — Oct 2025Data Synthesis Engineer · Intern 数据合成工程师 · 实习
- Led engine refactoring & HPC module architecture for embodied AI sim主导具身智能仿真引擎部分重构与HPC模块架构设计
- Built multi-camera parallel RT renderer on warp-lang + OptiX 8实现基于warp-lang与OptiX 8的多相机并行实时光追渲染器
- Developed DexGen — asset generation tool for embodied AI参与具身智能资产内容生成工具DexGen的系统设计与实现
Beijing Aerospace Measurement & Control 北京航天测控技术有限公司
Nov 2023 — Mar 2024Engine R&D Engineer · Intern 引擎研发工程师 · 实习
- Compressed 3DGS model to 1/10 size while improving high-freq spatial representation将3DGS数据压缩为原方法的1/10，优化高频空间信息表征
- Built toolchain for segmenting & extracting 3D objects from continuous video从视频序列中智能识别、分割并提取独立三维物体模型