Projects
Here are some Full Stack & Machine Learning projects I've worked on and am currently working on.
Full-Stack language learning SaaS -Amigolearn
• Developed a full-stack language learning SaaS using Next.js with Drizzle ORM serving a REST API with React as the frontend, ensuring a dynamic and responsive user interface.
• Implemented secure user authentication with Clerk and integrated AI-generated voices using Elevenlabs AI, enhancing both security and user engagement.
• Designed game mechanics including a points/XP system, hearts system, and leaderboard, fostering user motivation and engagement.
• Utilized Drizzle ORM and PostgreSQL (NeonDB) for efficient database management​
​
Full-Stack Recipe Web Application - Munchkin Menu
• Developed a web-based application allowing users to create, discover, comment on, and rate recipes, fostering a vibrant cooking community with interactive features
• Engineered a responsive front-end using ReactJS and integrated it with a scalable back-end built with Node.js and Express.js, ensuring seamless user interaction and efficient data processing
• Architected ORM with Sequelize to manage MySQL database interactions, designed and normalized tables to 3NF
New Recipe Recommender - GPT2 model with RAG
• Developed a chat-style Personalized Recommender Recipe System using a fine-tuned GPT-2 model on the Food.com dataset (~450,000 recipes).
• Applied natural language processing (NLP) and machine learning techniques to create a system that understands user dietary restrictions and taste preferences.
• Implemented Retrieval Augment Generation (RAG) with guardrails, utilizing 70% of the dataset to ensure high-quality, safe recipe recommendations.
​
​
Click here for the Github Page with Demo.
Image Classification with CNN model
In this project, I develop a Convolutional Neural Network (CNN) model using PyTorch to tackle the Intel Image Classification challenge. The goal is to accurately classify images into various natural scenes, such as mountains, forests, and cities. This project demonstrates my expertise in data handling, model training, and evaluation, showcasing CNNs' capabilities for complex image classification with PyTorch.
​
Click here for the Jupyter Notebook.
New Movie Recommendation
-
Designed and implemented a content-based filtering system for movie recommendations using cosine similarity on TF-IDF vectorized features.
-
Developed feature vectors representing movies by genres and descriptions, enhancing similarity measurement accuracy.
-
Iterated through a 10,000-movie dataset to identify and recommend movies with the highest similarity scores to user-selected favorites.
​
Personal Ticket System - GUI App
This is a personal GUI application for a data scientist or a software engineer. This intuitive GUI application facilitates the organized submission and tracking of new projects, tickets, and comments, enabling users to categorize, prioritize, and manage their workflows efficiently.
​
