Developed a multi-class classification system for finger-tap detection using deep learning. Implemented MediaPipe for real-time hand tracking and classified finger taps using an MLP (Multi-Layer Perceptron). Designed to monitor Parkinson's progression, tracking involuntary taps as a potential disease indicator.

Developed an AI-powered chatbot leveraging Retrieval-Augmented Generation (RAG) for context-aware responses. Used FAISS for efficient retrieval of knowledge chunks from structured sources. Optimized query embeddings to improve accuracy and reduce response latency. Personally trained this chatbot on a Python textbook, making it a study assistant for quick topic revisions.

Developed a Speech Emotion Recognition (SER) system using the Toronto Emotional Speech Set (TESS) dataset. Extracted key speech features using Mel-Frequency Cepstral Coefficients (MFCCs) and applied Principal Component Analysis (PCA) for dimensionality reduction. Trained multiple classifiers (SVM, Logistic Regression, KNN, and Random Forest) and identified SVM as the best-performing model with 91% accuracy. The system enables real-time speech emotion classification for AI-driven applications.

Developed a sentiment analysis pipeline to analyze customer reviews. Applied NLP techniques such as tokenization, stemming, stopword removal, and text vectorization (TF-IDF & BoW) for effective text processing. Built a machine learning-based sentiment classifier using Logistic Regression and Multinomial Naïve Bayes, achieving over 72% test accuracy. Designed data visualizations to explore sentiment distribution, providing insights for businesses to understand customer opinions.

Developed an AI-powered predictive maintenance system for aircraft turbofan engines using the NASA C-MAPSS dataset. The system forecasted potential engine failures, reducing unplanned maintenance downtime by 30% and improving operational efficiency. Leveraged machine learning models to predict maintenance needs with 95% accuracy, ensuring timely part replacements and minimizing unexpected failures.

Conducted a comprehensive customer churn analysis using a telecom dataset in SAS. Performed data cleaning, missing value imputation, feature engineering, and logistic regression modeling. Evaluated the model with ROC curves, precision, recall, and F1-score metrics. Implemented a manual 5-fold cross-validation pipeline to validate model robustness and predict customer churn with actionable insights.