Projects

Built a regression model to predict house prices using a housing dataset. Includes data cleaning, feature engineering, model training and evaluation. The trained model is exported as a .pkl file and integrated with a simple UI for predictions.

Developed a movie recommendation system in Python that suggests relevant movies based on user preferences and similarity logic. The project includes data processing, recommendation algorithms, and a simple interface to explore and test movie recommendations, along with clear documentation.

Built a Machine Learning–based Rock vs Mine classification system that predicts whether an object detected by SONAR signals is a Rock or a Mine. The model is trained on the SONAR dataset using supervised learning algorithms and deployed as an interactive Streamlit web application for real-time predictions.

Developed a Fake News Detection system using Natural Language Processing (NLP) and Machine Learning to classify news articles as real or fake. The project uses TF-IDF vectorization for text feature extraction and a Logistic Regression model trained on labeled news data, achieving an accuracy of 98.6%. The model is deployed as an interactive Streamlit web application for real-time news verification.

Developed a Machine Learning–based Diabetes Prediction system that predicts whether a person is diabetic or not based on medical input parameters. The project includes data preprocessing with feature scaling using StandardScaler and a supervised classification model trained on a diabetes dataset. The trained model is deployed as an interactive Streamlit web application for real-time predictions.

Built a fully functional Telegram-based food ordering bot using Python and python-telegram-bot v20. The bot allows users to browse a food menu, place orders step-by-step, select quantity, choose payment methods, and instantly notifies the owner with complete order details. Deployed on Render for 24×7 availability.

Developed an end-to-end Machine Learning system to predict customer churn based on service usage, contract type, and billing information. The project includes data preprocessing, exploratory data analysis, model training using Random Forest, and an interactive Streamlit web app for live predictions.

Developed a Machine Learning system to predict 30-day hospital readmission risk for diabetic patients. The project focuses on identifying high-risk patients before discharge using healthcare-safe evaluation metrics. Implemented techniques like class imbalance handling and threshold tuning, and deployed the model as an interactive Streamlit web application.

Built a Machine Learning–based credit card fraud detection system to identify fraudulent transactions from highly imbalanced real-world data. The project includes data preprocessing, exploratory data analysis, handling class imbalance using SMOTE, model training with Logistic Regression and Random Forest, and deployment as an interactive Streamlit web application with recall-focused evaluation.

Developed an end-to-end Machine Learning system to predict whether an e-commerce order will result in financial loss before shipping. The project uses real-world marketplace data, performs feature engineering, and applies a classification model with probability-based risk scoring. The system includes a production-ready FastAPI backend, business rule overrides, and a live frontend for real-time predictions.

Developed an end-to-end Machine Learning system to predict future medicine demand using historical healthcare data. The project is a full-stack application where trained ML models generate probability-based demand predictions through a FastAPI backend, consumed by a responsive frontend. This system helps pharmacies and healthcare planners proactively manage inventory and reduce medicine shortages.

Developed a Machine Learning–based predictive maintenance system to forecast potential machine failures using industrial sensor data. The solution analyzes equipment patterns and provides real-time predictions through an interactive Streamlit web application, helping reduce downtime and operational losses.

Developed an AI-powered crime analytics system that predicts the probability of arrest based on crime details such as type, location, time, and contextual factors. The project integrates a high-performance ML model with a FastAPI backend, a React frontend dashboard, and a Streamlit demo application to deliver real-time, data-driven insights for crime analysis.

Model Performance: Achieved ~0.90 ROC-AUC, enabling reliable prediction of arrest probability for real-world crime data scenarios.

Developed an end-to-end Machine Learning system to detect fraudulent financial transactions using highly imbalanced real-world data. The project includes advanced preprocessing, SMOTE-based class balancing, and threshold tuning to optimize precision-recall trade-offs. Deployed as an interactive Streamlit web application for real-time fraud prediction.

Built a content-based book recommendation system using TF-IDF and cosine similarity to suggest similar books based on user input. The system features an interactive Streamlit interface with real-time search, top recommendations, and book cover previews, delivering a fast and user-friendly experience.

Developed a modern Student Management System using Flask and SQLite with a clean cyber-style UI. The system enables secure admin authentication, student record management, fee tracking, and password reset functionality, providing a complete solution for managing academic data efficiently.

Developed a dynamic browser-based endless runner game with smooth animations, physics-based jumping, and real-time gameplay mechanics. The game features customizable assets, a chaser AI system, obstacle generation, and a modern cyberpunk UI, delivering an engaging and interactive gaming experience.