Projects
Rate My Professors Chatbot
Aug 2024
[Pinecone, Llama 3.1 API, Next.js, React, Material UI, Python]
Developed a chatbot that can help you find professors matching your needs by using (mock) Rate My Professors reviews, utilizing LLM (Large Language Model) API, RAG (Retrieval-Augmented Generation), and Vector Database.
CardFlip
Aug 2024
[Llama 3.1 API, Next.js, React, Material UI, Firebase, Clerk API, Stripe API]
Developed a smart AI-powered flashcard-generating platform supporting:
- Easy Text Input: Simply input your text and let our software do the rest. Creating flashcards has never been easier.
- Smart Flashcards: Our AI intelligently breaks down your text into concise flashcards. Perfect for studying!
- Accessible Anywhere: Access your flashcards from any device, at any time. Study on the go with ease!
- Dark Mode: and of course... Our favorite well-designed dark and light mode toggle!
DevOrbit
Aug 2024
[Llama 3.1 API, Next.js, React, Material UI, Firebase, Vercel]
Developed a Chatbot platform supporting:
- Smooth Chat Experience: the bots always keep a memory that belongs to only you two! Easily access and review past conversations with any bot, allowing you to revisit crucial advice, tips, and insights at any time
- Role-Based Conversation: Create new bots and customize them to fit your needs, from casual conversation partners to expert advisors, and switch between them effortlessly
- Dark Mode: Switch to dark mode for a more comfortable and focused experience, keeping you energized and productive.
DeEPsnap Model Construction
May–Jun 2024
[Python, Keras, PyTorch, NumPy, scikit-learn, matplotlib]
- Proposed and constructed a novel snapshot ensemble DNN(Deep Neural Network) using Python with Keras
- Boosted all testing metrics by 1~3% from baseline DNN for essential gene prediction
- Experimented with and outperformed GAT(Graph Attention Networks), Random Forest, AdaBoost, and SVM(support vector machine)
Image Semantic Segmentation Models Comparison
Feb–May 2024
[Python, PyTorch, torchvision, NumPy, matplotlib]
- Preprocessed and augmented 7000+ images and trimap labels of the Oxford-IIIT Pet dataset
- Implemented logistic regression, FCN(Fully Convolutional Network), and U-net using PyTorch; curated and adjusted source code from official repository for DeepLabV3+ model
- Trained, fine-tuned, and well-tested the models; compared performance by pixel accuracy and IoU on the test set
- Investigated the model structures and corresponding research papers to gain a deep insight into models for detailed comparison
- Crafted a detailed 18-page final report elucidating the dataset analysis, image processing methods, evaluation metrics, model structures, and interpretation of comparison results
DeepHE Model Construction
Feb–Aug 2020
[Python, Keras, torchvision, NumPy, scikit-learn, matplotlib]
- Constructed and applied Multilayer Perceptron with deep learning techniques such as dropout on biological multi-omics data to predict human essential genes
