---

👨‍💻 About Me

• 🎓 Junior at NEUQ (Northeastern University at Qinhuangdao) majoring in Computer Science
• 🔥 Passionate about system internals and large language models (LLMs)
• 💼 Seeking internship opportunities in Backend or AI Engineering
• 🌱 GPA: 2.84/5.0

---

🔗 Connect with Me

---

🔧 Skills

Core Competencies: Operating Systems | Machine Learning | Natural Language Processing (NLP)

---

🚀 Featured Projects

📂 OS Kernel Simulator — High Concurrency File System

Tech Stack: C++20 mmap msync Message Queue Semaphore Priority Scheduling FAT LRU

• Storage Layer: Designed DiskManager using POSIX mmap/msync to map 64KB local file into 1024 × 64B "disk blocks" for physical disk simulation
• File System Layer: Implemented FileSystem with FAT table, single-level directory, and LRU buffer pool (16 pages, 64B/page) for efficient file operations
• Process & Communication Layer: Built MessageQueue for priority-based inter-process communication; ProcessScheduler using std::counting_semaphore for CPU resource control
• Visualization: MonitorPanel for real-time FAT table, buffer pool, and disk usage monitoring; block-level hex + ASCII dual-view display
• Optimization: I/O efficiency improved by ~40% using LRU Buffer Pool and dirty page write-back; race conditions solved via semaphores and mutex locks

View on GitHub →

🤖 AI Companion — Roleplay & Memory System

Tech Stack: Python Streamlit DeepSeek API JSON

• Built an AI chatbot featuring persistent memory and emotional roleplay capabilities
• Designed a dynamic System Prompt mechanism for flexible personality injection
• Implemented CRUD operations for chat history using local JSON storage
• Achieved millisecond-level streaming response and seamless context switching

View on GitHub →

🔍 Jigsaw Toxicity Prediction — NLP Classification

Tech Stack: Python PyTorch NLP Transformer

• Built a text toxicity classifier using Transformer-based models
• Processed and analyzed large-scale NLP datasets for sentiment analysis
• Implemented multi-label classification for various toxicity types
• Achieved competitive performance on Kaggle Jigsaw competition benchmarks

View on GitHub →

🔥 PyTorch Deep Learning Lab

Tech Stack: Python PyTorch

• Experimented with various neural network architectures and training techniques
• Implemented custom loss functions and optimization strategies
• Explored model optimization techniques including quantization and pruning

View on GitHub →

📊 EM Classification App — Streamlit Visualization

Tech Stack: Python Streamlit Expectation-Maximization

• Built an interactive Expectation-Maximization algorithm visualization
• Implemented real-time clustering visualization with Streamlit
• Designed intuitive UI for parameter tuning and result exploration

View on GitHub →

---

🎓 Education

Northeastern University at Qinhuangdao | B.S. in Computer Science

Course	Score
Data Structures	84
Operating Systems Principles	83
Compilation Principles	83
Java	90

---

"Procrastination is the thief of time."