Projects
FinTech RAG Model
The FinTech RAG (Retrieval-Augmented Generation) model extracts and processes data from financial PDFs by identifying relevant tables and text. It handles two types of queries: span-based (for text extraction) and arithmetic (for solving mathematical problems). The model then feeds this data into a large language model (LLM) to generate precise answers, enhancing data analysis in the financial domain.
Image Captioning with Attention Mechanism
This project is a deep learning-based solution for generating captions for images. It uses a Convolutional Neural Network (CNN) as an encoder with transfer learning, a Recurrent Neural Network (RNN) as a decoder, and an Attention Mechanism to dynamically focus on relevant parts of the image during caption generation. The implementation is done using PyTorch.
llm-PDF-Chat
This application is a conversational assistant built with Streamlit, designed to process PDF documents, extract text, and interact with users through a language model. It leverages langchain, FAISS for vector storage, and OpenAI's embedding and language models to enable intelligent responses based on the content of uploaded PDFs.
Stock Price Prediction
This project predicts Google stock prices using historical data and machine learning techniques. It involves data preprocessing, feature engineering, and model development using LSTM (Long Short-Term Memory) networks to capture temporal dependencies for accurate predictions. The implementation includes visualizations and performance evaluation to ensure reliable forecasting.
Customer Segmentation
Enhanced customer segmentation by applying KMeans clustering to a retail dataset, identifying distinct customer groups, streamlining data preprocessing through feature engineering, and categorizing customers into actionable segments, resulting in improved model performance, targeted marketing strategies, and data-driven decision-making to boost customer retention and engagement.
Text Summarizer
A text summarization project using Hugging Face transformers, specifically the Pegasus model, which focuses on processing dialogue datasets. The project includes comprehensive pipeline stages of data ingestion, validation, transformation, model training, and evaluation, with robust error handling and logging mechanisms.
