Projects: Gesture Based Mouse Using Python Description: Hand gestures translated to cursor actions for contactless computer interaction. Created a Virtual mouse that uses hand gestures for controlling the cursor using Python, OpenCV and MediaPipe. I enabled touchless interactions by mapping gestures to mouse movements and clicks. I also focused on real-time processing and accuracy across diverse environments. Key contribution:  Built a basic gesture recognition system using Python, OpenCV and MediaPipe to control mouse movements.  Processed live Webcam feed to track hand landmarks and map finger gestures to cursor actions.  Mapped gestures to mouse operations like move, click, drag and scroll using Geometric Calculations  Applied Smoothing Techniques using NumPy to ensure cursor stability and reduce the jitters.  Tuned the gesture recognition to perform reliably by using different lights and background conditions.  Packaged the solutions into a self-contained Python Script with minimal dependencies for easy use. PredictiCare: Health Predictions Using ML Description: Machine learning models for predicting health issues to support proactive care. I have built ML models to predict potential health issues based on user data and medical history. I implemented classification algorithms like Neural Networks, Classification trees, Naïve Bayes, Random Forest, XgBoost, ADABoost and Logistic Regression for early diagnosis support with high accuracy. I aimed to assist for taking proactive decisions through Predictive Analytics. Key contributions:  Built a supervised Machine Learning model using Scikit-learn to predict health conditions based on patient data.  Cleaned and prepared the dataset using Pandas which includes missing value handling, label encoding and feature scaling.  Trained and compared models such as Logistic Regression, Random Forest and XGBoost to optimize the performance.  Achieved approximately 85% accuracy by using Hyperparameter tunning and validating results using Cross-validation technique.  Used Matplotlib and Seaborn for Exploratory Data Analysis and results visualization.  Interpreted model outputs using feature importance to identify key health risk indicators.  Documented the pipeline in Jupyter Notebook and shared results with a simulated clinical dashboard view. Fraud Shield Using Deep Learning Description: System designed to detect fraudulent activities in real time using deep learning models. I had built a real time fraud detection system using Deep Learning on the transactional data. I leveraged LSTM and dense Neural Networks to identify the suspicious patterns in the financial data. I also improved fraud recall significantly by enhancing financial system security. Key contributions:  Designed a fraud detection system using TensorFlow and Keras by leveraging LSTM to capture sequential patterns.  Pre-processed synthetic transactions data using Pandas which handles the outliers and transforming categorical variables.  Structured time-series input data for LSTM by reshaping transactional histories for each user.  Handled imbalanced classes using class weights and evaluated model performance with Precision, Recall and AUC.  Achieved over 90% fraud Recall by minimising False Negatives in testing situations.  Visualised Confusion Matrix and ROC Curve using Matplotlib for fraud vs non-fraud interpretation.