A Python tool for analyzing faces in images, featuring gaze detection, eye state classification, and emotion recognition using deep learning models.

An image classifier model trained on the CIFAR-10 dataset; the repository contains a custom CNN model and a pretrained VGG16 model for accurate image classification tasks.

A project featuring two Torch models, a fully connected neural network (FNN) and a convolutional neural network (CNN). Both trained on the MNIST dataset for handwritten digit recognition.

Fine-tuned the Florence Vision-Language Model (VLM) for an image-to-text task, where it translates visual meter readings into descriptive text outputs for accurate and automated data interpretation.

A generative AI app that converts natural language questions into SQL queries and retrieves results from a database. Built with Streamlit and powered by LLMs, it makes database access easy and intuitive for non-technical users.

A simple web application built with Flask that allows users to register, log in, and manage their favorite websites by securely storing URLs in a SQLite database.

A deep learning project built with TensorFlow that uses a U-Net model for brain tumor segmentation on the BraTS dataset. It accurately identifies and segments tumor regions in MRI scans

A multi-task model for human action recognition (40 classes) and person count (binary) using TensorFlow and Keras, experimenting with pretrained models, fine-tuning, and dual output layer configurations

A social networking app built with Django, allowing users to register, log in, create posts, and follow/unfollow others. The frontend uses Volt Bootstrap 5 for a modern UI, and future updates will include AI-powered chat features to enhance the user experience.

A web application for tracking personal finances, integrating Fivetran, Google Sheets, BigQuery, and Looker Studio. It allows users to input expenses and income data, visualize financial trends, and gain actionable insights through a dynamic BI dashboard.

An automated image processing tool developed in Python, using CLAHE for image enhancement and color thresholding (using CV2) in the HSV color space for precise image segmentation. The tool is deployed with Flask and an HTML frontend for easy user interaction.

A deep learning-based solution using the Segment Anything Model (SAM) for image segmentation and VGG16 for disease classification. The workflow incorporates HSV color thresholds and OpenCV operations to enhance segmentation, achieving high accuracy on both segmented and non-segmented datasets for early tomato plant disease detection.

Automating sentiment analysis for YouTube comments using Python, Transformers, and the YouTube API. The tool fetches comments from any YouTube video, analyzes sentiments using pre-trained models, and visualizes the results, helping uncover insights from user feedback.

A Power BI dashboard built using OLIST e-commerce data to provide actionable insights for sales, customer satisfaction, and delivery performance. The dashboard includes advanced visualizations, DAX formulas, and forecasting, helping businesses make data-driven decisions.

A tutorial in designing clearer and more effective Power BI dashboards by correcting common visualization mistakes and improving layout, chart selection, and color schemes.

A project on managing large datasets with MySQL using the Adventure Works database. Tasks include cleaning data, setting up relational schemas, importing CSVs via Python automation, running complex SQL queries, and analyzing sales, customer behavior, and product profits.