PSU HelpBot With Langchain Integration

Project Description: This question-answering system helps students quickly find the information they need about PSU's graduate programs, making it easier for them to make decisions. It uses advanced AI models like OpenAI's GPT-3.5 and web scraping to provide accurate and fast results.

In this project, I developed a question-answering system to help students efficiently access information about PSU's graduate programs. Utilizing Langchain and OpenAI's advanced language models (including GPT-3.5-turbo-instruct and Davinci-002), I implemented web scraping techniques to gather targeted data from PSU's graduate and computer science department websites. The system employs a Retrieval Augmented Generation (RAG) approach, utilizing similarity search and MMR techniques within the RetrievalQAchain to ensure accurate information retrieval. The final solution was fine-tuned with BERTScore and RAGAS metrics, achieving 87% accuracy and providing a user-friendly experience.

National Park Chatbot using dialogflow and flask

Project Description: This Flask application serves as a webhook for Dialogflow, facilitating interaction with a National Park Chatbot. Integrated with features such as activities, alerts, and weather updates, the chatbot enhances user experience in accessing information about national parks.

Leveraged advanced design principles to enhance user engagement and experience, building upon the foundation laid by the initial project. Incorporated features such as Welcome Intent, Activity and Alert Information retrieval, Weather Information, and State-based Park Name listing, utilizing open-source Weather APIs and the National Park API to fetch data dynamically. Employed a modern tech stack including Flask, Dialogflow, and React to ensure scalability and maintainability. Deployed the application on Render.com, seamlessly integrating it with Dialogflow Messenger and Dialogflow Web Demo for voice assistance.

Advanced NPL for classify the consumer complaints

Project Description: This project focuses on developing a Natural Language Processing (NLP) model to streamline the categorization of consumer complaints received by the Consumer Financial Protection Bureau (CFPB).Consumer complaint narratives serve as inputs, and the model classifies them into five consolidated product classes: Credit Reporting, Debt Collection, Mortgages and Loans, Credit Cards, and Retail Banking.

Developed and implemented a highly accurate NLP system, achieving an impressive 84% accuracy rate using XGBoost on a dataset comprising 625,545 consumer complaints. Demonstrated proficiency in various stages of the data science pipeline, including data preprocessing, model evaluation, and ethical considerations, aligning with industry best practices. Leveraged advanced pre-trained language models such as BERT, GPT2 resulting in an 82% accuracy rate in categorizing consumer complaints, showcasing adaptability and expertise in utilizing state-of-the-art techniques. Created Flask web application to provide a user-friendly interface to classify the consumer complaints easily.

National Park Dashboard

Project Description: Park Dashboard uses the national park API for detailed stats and info on different parks. Users can search by name to find current weather, alerts, activities, trails, photos, fees, and contacts. We also offer charts for park counts by state, alerts, activities, amenities, and overall national park count.

My responsibilities encompassed designing and implementing several key elements. These included crafting the navbar section, developing informative card components featuring details like weather forecasts, park activities, recommended attractions, campgrounds, and more. Additionally, I constructed line charts to showcase the 5-day weather forecast, generated bar charts for state vs. national park comparisons, as well as state vs. generic parks analyses. Furthermore, I designed 3D pie charts to visualize alerts categorized across all parks and national parks. Additionally, I integrated pop-up modal components to enhance user interaction.

Digital Souvenir - Travel Memories

Project Purpose: People always buy souvenirs whenever they travel places. With memories now people will have the opportunity to store the souvenirs digitally with personalized images and thus share them with others.

My Responsibility: I focused on configuring and linking Firebase, enabling seamless image and metadata storage. I also developed essential features like updates, deletions, and searches. Utilizing the Google Search API's autocomplete, I enhanced location searches. The front-end came alive with React Bootstrap, where I crafted modals, forms, and engaging search cards.

Book Recommendation System

The purpose of this project is to build a model which recommend books to readers based on popularity parameters and collaborative approach.

This website offers personalized book recommendations in five categories: similar books, authors, publishers, places, and years. The backend, developed in Python, handles the data and recommendation logic, while the frontend, created with Flask, showcases books grouped by these categories. Users enjoy customized book suggestions tailored to their preferences.

Website to collect ASL video dataset

This project aims to bridge communication gaps, enabling seamless interaction between the hearing and non-hearing communities. We're crafting a robust framework that instantly converts sign language into text or voice

To achieve this objective, the project necessitates efficient datasets. To meet this goal, I've designed a website to collect vital datasets efficiently.

This website is built upon the foundation of ReactJS, HTML/CSS, and Firebase technologies. Its primary function is to gather essential data encompassing 50 distinct categories of American Sign Language (ASL) and Indian Sign Language (ISL)

Rubik’s Cube Solver Using Python

Project Purpose: Design an application to solve the Rubik’s Cube using a camera

I've built a program that takes pictures of all the cube's sides using the camera, and then it guides users through the steps to solve it. To make this possible, I've utilized a Python module called Kociemba, which has smart algorithms to help with solving. This project brings together the camera, programming, and clever algorithms to make solving a Rubik's Cube easier and more enjoyable.