Beyond Stars
Beyond Stars is an interactive web app that lets users explore the night sky as seen from exoplanets. Using data from NASA’s Gaia star catalog, the app provides a unique 3D view of stars and constellations from various exoplanetary perspectives. It also features an educational gaming mode to engage users in discovering celestial objects.
Tech Stack
- Languages: Python, JavaScript, HTML/CSS
- Frameworks & Libraries: Three.js, WebGL
- Tools: Figma (for UI/UX design)
- Data Source: ESA Gaia DR3 Star Catalog
UA Course Compass
UA Course Compass is a web-based software application designed to help students pursuing a Bachelor of Science in Information Science at the University of Arizona effectively manage their 4-year course plan. The platform provides personalized course recommendations, a four-year planning tool, interest surveys, and email reminders to streamline the academic planning process, leveraging machine learning and data scraping technologies to enhance the student experience.
Tech Stack
- Languages: HTML, CSS, JavaScript
- Frameworks & Libraries: Node.js, Selenium
- Database: PostgreSQL
- Machine Learning: NLP algorithm for course recommendations
- Other Tools: Email API for notifications, Data scraping from university course catalogs
Autonomous Deep Space Exploration
This project focuses on enabling deep space exploration using innovative system-of-systems architecture with small spacecraft inspectors. Leveraging machine learning for real-time decision-making, the project addresses the challenges of long-duration space missions, including navigation and communication limitations in harsh environments.
Tech Stack
- Languages: Python
- Frameworks: Machine Learning Algorithms for autonomous operation
- Tools: Synthetic Data Generation, Hyperparameter Tuning
- System Architecture: Multi-layered inspector satellite design
Linear Regression and Cross Validation
This project focuses on implementing linear regression using matrix normal equations and performing cross-validation to assess model performance. The objective is to fit polynomial models to datasets and interpret regression outcomes. Python is used for building regression models, generating cross-validation results, and producing plots for data visualization.
Tech Stack
- Languages: Python
- Libraries: NumPy, Matplotlib
- Tools: Cross-Validation, Linear Regression, Polynomial Regression
Agentic Customer Support - Automating Customer Service Workflows
This project focuses on building an intelligent customer support system that leverages LangGraph and OpenAI’s GPT API. The system automates query categorization, sentiment analysis, and response generation, while dynamically visualizing workflows for improved debugging and clarity. It includes mechanisms for escalating unresolved issues to human agents, ensuring a human-centric approach to customer service.
Tech Stack
- Languages: Python
- Libraries: LangGraph
- APIs: OpenAI GPT API
- Tools: Workflow Visualization, Sentiment Analysis
Mental Health Chatbot for Counselors – AI-Driven Mental Health Support
This project focuses on fine-tuning a LLaMA-3 model for mental health applications using Unsloth, an optimized framework for training large language models efficiently. It implements QLoRA for memory-efficient training and is deployed on Hugging Face for real-time interaction using HumeAI. The system provides counselors with a reliable, empathetic tool to support mental health care.
Tech Stack
- Languages: Python
- Libraries: QLoRA, Unsloth
- APIs: HumeAI
- Tools: Hugging Face
OutfitEquation – Agent-Based Fashion Recommendation System
This project focuses on developing an agent-based fashion recommendation system that leverages the Polyvore dataset and the PinAI API for data retrieval. It delivers personalized outfit suggestions based on skin tone, budget, and past inspirations, offering real-time style-based recommendations for users.
Tech Stack
- Languages: Python
- Datasets: Polyvore
- APIs: PinAI API
- Tools: Agent-based Modeling
Next Project
Stay tuned for an exciting new project!