I'm an undergraduate student at Peking University, studying mathematics and building things around machine learning, web products, and small games.

Currently, I care most about understanding how models learn — from symbolic grokking experiments to LLM loss curve prediction.

Research / ML

• DL-final: loss curve prediction under different learning-rate schedules.
• Modelmid: math-solution authorship classification across human and LLM-generated answers.
• ML-final: small Transformer experiments on modular arithmetic, group operations, and grokking.

Web / Product

• Academic-Department-Website: official website for the Academic Department of SMS Student Union, PKU.
• SmartTime: an AI-assisted task and calendar management tool.
• wjjpku.github.io: personal website.

Fun stuff

• UncleDucky: a fictional chat-style management game.
• Candle of Faith: a Godot side-scrolling game made for a course project.
• PKUsmsExam: an archive of past exams for PKU math courses.

Python / PyTorch / TypeScript / React / Vite / Tailwind CSS / Supabase / LaTeX / Godot

research:  LLM loss curves, learning-rate schedules, grokking
building:  academic tools, AI-assisted productivity apps, small games
taste:     math first, product second, decoration last