// connected to the wired

Zhuoran Wang

王 卓 然

Building at the intersection of generative AI and digital entertainment.
Interested in AI VTubers, diffusion models, and the culture of the Wired.

↓ scroll
// 01. about

Who Am I

Zhuoran Wang

I'm a builder and researcher who believes the most interesting things happen at the edges — where mainstream attention hasn't reached, where there's room for genuine creation.

My current obsession is bringing AI characters to life. I'm building an AI VTuber from scratch, exploring how language models, voice synthesis, and real-time animation can converge to create something that feels alive and has a soul.

Academically, I work on generative models and efficient model inference & training. Outside of research, I spend most of my time thinking about cyberculture, anime, and the strange places where technology and subculture collide.

AIGC VTuber Diffusion Models LLM Inference Voice Synthesis Anime Cyberculture Open Source
// 02. blog

Transmissions

2026.03
Innovation Happens at the Edges
Why the most important breakthroughs in computing always came from the margins — and what that means for AI.
Cyberculture
2026.03
The Death and Rebirth of Internet Spirit
1990s internet culture, the commercialization of the web, and where genuine creation lives today.
Essay
2026.02
Open Source AI VTuber Landscape: A Technical Survey
Comparing Project Airi, Neuro-sama's architecture, and the emerging ecosystem of AI-driven virtual characters.
Technical
→ All posts
// 03. projects

Building

● In Progress
AI VTuber
An end-to-end AI-driven virtual character capable of real-time voice dialogue, emotional expression, and Live2D animation. Built with LLM backbone, neural TTS, and real-time lip sync. Inspired by Neuro-sama and the vision of AI characters with genuine personality.
DeepSeek Kokoro TTS faster-whisper Live2D Python
● Open Source
SGLang Contributor
Contributing to SGLang, a structured generation language for LLM inference optimization. Documentation improvements and community engagement.
SGLang LLM Inference Open Source
// 04. academic

Credentials

Publications

Towards Dynamic KV-Cache Compression: Fine-Grained Evaluation of Key and Value Ranks in LLMs
Jian Chen*, Zhuoran Wang*, Jiayu Qin, Ming Li, Meng Wang, Changyou Chen, Yin Chen, Qizhen Weng, Yirui Liu
NeurIPS Workshop, 2025 (* equal contribution)

Education

MSc in Computer Engineering
Delft University of Technology, The Netherlands
Sep 2023 – Nov 2025
BSc in Robotics and Intelligent Devices
National University of Ireland, Maynooth & Fuzhou University (Dual Degree)
Sep 2019 – Jul 2023
First-class Honors, GPA 3.8/4.0

Experience

Research Intern — TeleAI, China Telecom
Optimizing verl and vLLM for multi-modal video RLHF training
Jun 2025 – Present
Master Thesis — TU Delft, Embedded Systems
Efficient Large Language Model Inference
Feb 2025 – Present

Service

Reviewer — AAAI Undergraduate Consortium, 2026
Student Volunteer — ASPLOS / EuroSys 2025, Rotterdam