Projects:

Pharmaceutical management system:

• Tech Stack:JavaScript, CSS, HTML, Java, Shiro
• Led the overall project architecture design and technology selection, implementing a decoupled front-end (Vue.js) and back-end (Spring Boot) solution.
• Designed and implemented the security framework using Apache Shiro, delivering user authentication, session management, and Role-Based Access Control (RBAC).
• Developed and managed the core business modules, including backend APIs for drug information, supplier management, inventory (inbound/outbound), and sales statistics.
• GitHub Link

AI-Assisted Interview Platform:

• Tech Stack: Python, PyTorch, QLoRA, Ollama, HTML, Vue 3
• Led the creation and open-sourcing of the world’s largest Chinese interview dataset (150k+ entries) by scraping tech communities and generating high-quality Q&A pairs using multiple LLMs (GPT-4o, DeepSeek), and published it on Hugging Face.
• Directed the technology selection and fine-tuning of the AI interviewer LLM. Employed QLoRA (Unsloth) on a DeepSeek base model, training it on the custom dataset to specialize in generating interview questions, follow-ups, and automated scoring.
• Designed and deployed the project’s full-stack architecture: built the frontend application (Vue 3) integrating resume editing, AI polishing, and mock interviews; and deployed the fine-tuned model as an HTTP API service using Ollama.
• GitHub Link
• Dataset Link

Guess the numbers game:

• Tech Stack: Java, Java Swing, Java AWT, Java I/O
• Designed and implemented the complete game logic for a “Bulls and Cows” (nAnB) guessing game, building the entire multi-window graphical user interface with Java Swing.
• Developed a persistent user authentication and scoring system, managing user registration, login, and score ranking by reading and writing to local .txt files (Message.txt, Rank.txt).
• Enhanced the user experience by creating custom, reusable Java Swing components (CustomButton, CustomTextField) that support background images.
• GitHub Link

Solitaire:

• Tech Stack: Java, Java Swing, Java AWT (Graphics, Event Handling), Java I/O
• Developed the complete game logic for Klondike Solitaire, employing a strong Object-Oriented design with a CardPile class hierarchy (Deck, Discard, Table, Suit) to manage game rules and state.
• Built a fully interactive GUI using Java Swing, handling complex mouse drag-and-drop events for selecting and moving single or stacked cards.
• GitHub Link

SNP-SNP_Prediction:

• Tech Stack: Python, pandas, scikit-learn, numpy, scipy
• Implemented a data analysis pipeline including Chi-Squared tests, T-tests, and mutual information algorithms to evaluate SNP-phenotype associations.
• Validated findings using a scikit-learn Logistic Regression model with cross-validation; achieved a power value of 1.0 in both biallelic and triallelic models, indicating 100% accuracy in identifying all target SNP loci.
• GitHub Link

Computational Optimization of Many Body Problems:

• Project Goal: To optimize an N-body simulation program by addressing its O(N^2) computational complexity.
• Core Technologies: Utilized OpenMP for multi-core parallelization combined with SIMD (AVX/AVX512) instructions for vectorized computation.
• Optimizations & Results: Enhanced memory efficiency through SoA layout and 64-byte alignment, achieving an 11.437x speedup on 32 cores while maintaining high computational precision.
• GitHub Link