Projects
Projects
F1GPT is a Retrieval-Augmented Generation (RAG) chatbot I built to learn Formula One as a complete beginner. The project combines web scraping of F1 sources, Gemini text-embedding-004 for embeddings, Astra DB for vector storage, and Gemini 2.0 Flash for chat generation. Built with Next.js and TypeScript, it features a Formula One-inspired interface with glass morphism effects and racing aesthetics. The cost-effective architecture uses free tiers throughout, making it accessible for student developers. F1GPT helps newcomers understand complex F1 concepts through conversational AI, transforming scattered information into accessible knowledge.
Predicting stock market prices accurately remains a significant challenge due to the dynamic nature and complexity of financial markets. Recent advancements in machine learning and deep learning methodologies have provided promising frameworks for enhancing the accuracy of stock market forecasts. This project explores various statistical and deep learning approaches, including ARIMA, SARIMA, LSTM, GRU, and an ensemble model combining LSTM and GRU architectures, to forecast stock prices using historical NIFTY 100 index data.
This project uses Convolutional Neural Networks (CNNs) for early plant disease detection from leaf images, aiming to improve agricultural practices. Our custom CNN model is trained on a diverse dataset to achieve high accuracy and effective disease classification.
The project includes a detailed evaluation with visualizations of training performance and model predictions. It introduces a novel CNN approach and a unique dataset for enhanced plant disease detection and management.
This project involves developing an AI-driven web scraper that extracts and structures information from websites based on user inputs. The scraper utilizes advanced Natural Language Processing (NLP) capabilities through a fine-tuned LLaMA 3.1 model, specifically fine-tuned using the Supervised Fine-Tuning (SFT) technique with LoRA and the Unsloth library. The application features a user-friendly Streamlit interface, allowing users to enter website URLs, specify content extraction parameters, and receive structured data outputs.
In collaboration with Dr. Prakash Kalingrao Aithal and Vishal, we are conducting an in-depth study on the "Impact of Global Events on Financial Markets." Our research aims to analyze how various global phenomena influence financial markets using advanced financial models. We are implementing three key models: the Capital Asset Pricing Model (CAPM), Arbitrage Pricing Theory (APT) Model, and the Markov Model. By comparing these established models with our own, we aim to gain a comprehensive understanding of market behavior in response to global events
Fisheye lenses are popular for capturing wide fields of view, often ranging from 120 to 180 degrees, making them ideal for applications like image-based rendering, single view metrology, 3D reconstruction, and robot navigation.
However, these lenses introduce significant distortions, causing objects to appear curved, especially at the edges of the image.
Our project presents the "Midpoint Circle Algorithm for Real-Time Fish-Eye Distortion Rectification" and "Hemi-Cylindrical Unwrapping Algorithm" a novel approach to correcting fisheye distortion.
Investing in mutual funds through a Systematic Investment Plan (SIP) is a popular method among investors for wealth creation. The objective of this project is to analyze the impact of different SIP dates on the returns of mutual funds. By investigating the variations in returns based on the SIP investment date, investors can make more informed decisions and potentially enhance their investment returns.
This project was for personal use. Any information provided in this project should not be construed as financial advice.
This project presents a comprehensive machine learning approach to predict customer churn in a telecommunications company. Utilizing a rich dataset, we employ various data pre-processing techniques, including encoding of categorical variables, feature scaling, and handling missing values. The study explores multiple machine learning algorithms like Logistic Regression, Random Forest, and Support Vector Machines, assessing their performance
The Health Android app is a comprehensive wellness application designed to support users in managing their health and lifestyle. It offers features such as step tracking, BMI calculation, dietary guidance, nicotine tracking, and personalized assessments to help users stay informed and make healthier choices. Through interactive and easy-to-navigate screens, the app promotes daily engagement in health-related activities, making wellness accessible and actionable.
FitnFit is a comprehensive online platform designed to revolutionize fitness tracking and health management. The website focuses on delivering an intuitive user experience and robust functionality and facilitates seamless tracking of workouts, nutrition, and overall wellness.