Full-stack development building on AnkiDroid’s open-source codebase, with frontend in Kotlin, backend SQLite database management, and integration of OpenAI’s Realtime API (real-time streaming LLM interactions), speech-to-text, and text-to-speech APIs
A self-initiated group project designed to help users practice language skills with an LLM through adaptive learning that adjusts to their proficiency level
A solo web application developed in HTML, CSS, and JavaScript to fetch and display random cat images via REST API, with user-specified tags, text, font styles, and image filters for personalized image generation
A BERT transformer model fine-tuned using PyTorch and Hugging Face Transformers for hate speech detection, achieving 82% accuracy on test data and 0.23 cross-entropy loss and evaluated performance with scikit-learn
Natural language processing (NLP) implemented for text preprocessing and tokenization pipelines to prepare large-scale tweet datasets for training