Fun. Innovation. Collaboration.
I'm passionate about applying my skills to real-world challenges through side projects. Here, you'll find the 18 of my best works in the fields of machine learning, computer science, automation and more, showcasing my ability to innovate and solve complex problems. Some projects are still active, others have been discontinued.
Featured Projects
Ensemble-UQ-LLM
Engineered a scalable framework for training ensemble models leveraging LLaMA architecture, incorporating Low-Rank Adaptation (LoRA) for efficient fine-tuning, and providing advanced uncertainty metrics including variance, entropy, and mutual information for binary classification tasks.
Python, LLaMA, LoRA, Hugging Face, YAML, PyTorch, CUDA, BoolQ, Uncertainty Metrics, Shell Scripting
MastoGraph - Mastodon
Developed a scalable social media analysis framework to perform real-time data crawling, NLP-based toxicity classification, and network analysis, utilizing advanced graph algorithms like PageRank and modularity for community detection and influence mapping.
Python, Mastodon API, Llama3-8B, LangChain, NetworkX, PyTorch, Matplotlib, Seaborn, Community-Louvain
x-of-Thought Reasoning
Designed an advanced reasoning visualization platform utilizing Chain of Thought (CoT) and Graph of Thought (GoT) models, featuring interactive visualizations, dynamic node exploration, and sentiment analysis, seamlessly integrated through a Streamlit-based web interface.
Python, Streamlit, NetworkX, LangChain, PyTorch, OpenCV, Pandas, Pytorch Audio, Scikit-Learn, Plotly, Sentiment Analysis, Dynamic Graphs
OpenForensics-DeepFake
Developed a deepfake detection system with 98.59% accuracy leveraging Microsoft’s Swin Transformer, featuring a robust evaluation pipeline and advanced data augmentation techniques for enhanced reliability in forensic applications.
Python, PyTorch, Transformers, Timm, HuggingFace, Lightning, Albumentations, OpenCV, Model Evaluation, Data Augmentation, Computer Vision