YuHang Wang

Ayanami1314

B.Eng. Student @ SJTU-SE

About Me

I am an undergraduate student at Shanghai Jiao Tong University (SJTU), majoring in Software Engineering. I am passionate about Large Language Models (pre-training, post-training, code agents) and AI for Software Engineering.

Currently, I am an AI research intern at Xiaohongshu Dots, working on pretrain data (code) and midtrain. Previously, I interned at Xiaomi MiMo, ByteDance (Douyin), Metabit Trading, and BondiTech. I am an active contributor to open-source projects and a strong believer in the power of open-source culture.

Research Interests

🧠

LLM

Pre-training, Post-training, RL

🤖

AI for SE

Code Agents, Software Engineering Automation

⚙️

Systems

Operating Systems, Agent Native Infra

Experience

Xiaohongshu - Dots2026.9 - Present

Pretrain Data Researcher

Working on pretrain data (code) and midtrain for large language models.

Mentored by Guanting Chen

Xiaomi - MiMo2026.3-2026.8

AI Research Intern

Worked on LLM pre-training, post-training, and code agent development.

Mentored by Hongshen Xu & Linghao Zhang

ByteDance - Douyin2025.10 - 2026.2

AI Research Intern

Research on coding agents and code repository pruning for large codebases, resulting in the SWE-Pruner paper.

Mentored by Shilin He & Siyu Ye

Metabit Trading2025.7 - 2025.9

Gen-AI Intern

Participated in LLM deployment and serving optimization for code agent workloads.

BondiTech2025.2 - 2025.6

AI Research Intern

Optimized RAG pipelines & AI workflows.

Selected Open-Source Projects

SWE-Pruner

First Author

Context pruning tool for LLM-based software engineering tasks. Helps reduce context size for coding agents.

GitHub318 stars

SWE-Pruner Pro

First Author

LLM-natived context pruning for tool calling. Read its mind, not its output.

GitHub26 stars

SWE-Explore Bench

Co-First Author

Benchmarking How Coding Agents Explore Repositories.

GitHub42 stars

FastContext

Contributor

Training Efficient Repository Explorer for Coding Agents.

GitHub

Chcore / OS-Course-Lab

Contributor

Operating system course lab series designed by IPADS lab at SJTU, covering ICS/CSAPP and microkernel OS implementation.

GitHub555 stars

Jcourse: A Course Selection Community for SJTU

Core Contributor

Course review and selection community for SJTU students, built with Go.

GitHub

ProPilot

Owner

An AI-powered coding assistant / agent project exploring LLM-driven development workflows.

GitHub

Education

Shanghai Jiao Tong University

B.Eng. in Software Engineering | SJTU-SE