☕️I am Xu Huang, a third-year undergraduate student with a strong competitive programming background, including an ACM-ICPC Gold Medal, NOI Silver Medal, and 6th place nationally in CCSP (CCF Collegiate System and Programming Competition). I was selected for Tencent’s Spark Program (accepting only 50 participants globally each year) and served as a key technical developer at a startup backed by Miracle Plus Alternative Investment (formerly YC China). My research interests focus on 3D scene/video understanding and generation, world models, and multimodal foundation models. I have previous experience working on AIGC for 3D applications. I am actively seeking PhD positions and am passionate about making impactful contributions to the field of AI.
我叫黄旭, 是一名大三本科生，我在算法竞赛方面取得了优异成绩，包括ACM-ICPC金牌、NOI银牌，以及在CCSP（CCF中国大学生系统与程序设计竞赛）中获得全国第六名。我曾入选腾讯星火计划（每年全球仅招收50人），并在奇绩创坛（原YC中国）投资的初创企业担任核心技术开发。目前我主要研究兴趣集中在3D场景/视频的理解与生成、世界模型（World Model）以及多模态基础模型方向。此前，我在AIGC for 3D方向有过一些研究工作。我正在寻找PhD职位，希望能够为改变世界贡献自己的力量。
Experience
AI Research Intern
Tencent IEG Game AI CenterLed an AI 3D layout generation project addressing limited 3D scene data by distilling scene arrangement knowledge from 2D vision models. Developed a coarse-to-fine pipeline combining visual models with scene graph optimization, implementing iterative refinement through generated 3D scenes to achieve diverse layouts in novel 3D databases. Aiming for Siggraph submission.Research Intern
PKU Melon LabConducted 6-month research on various GPU parallel computing methods for quantum state generation and their efficiency impacts. Interested in ICML submission.Core Developer
ZhibeiTech- Led full stack development (backend-focused) for 3 projects, including an AI-powered educational content platform
- Developed an intelligent system that transforms diverse educational materials (PDFs, lecture recordings, PPTs, handwritten notes) into structured notes, mind maps, and podcasts
- Gained deep expertise in advanced AI techniques including Chain-of-Thought reasoning, prompt engineering, and Retrieval-Augmented Generation
- Latest project achieved $10000+ revenue within first month of launch
Research
Proposed a novel Mutual information based approach for RAG retrieval compression, significantly outperforming SOTA with > 25% improvement in ACL submission. url: https://arxiv.org/abs/2408.10497v1AI Research Intern
Tencent- Reproduced Vector Fusion paper for text-to-SVG generation using Score Distillation Sampling (SDS) and diffrentiable rendering
- Implemented automatic vector graphics generation pipeline with diffusion models
- Optimized Bezier curve parameters for high-quality SVG output
- url: https://github.com/ydove0324/vector_fusion
Education
BSc Computer Science and Technology
Beijing University of Posts and TelecommunicationsGPA: 93.15/100 (Rank: 2/377)