AI / NLP Predictive Models

Technical Contributions

• Implemented ANN, CNN, RNN, regression, clustering, and decision tree models.
• Built and deployed chatbots using Amazon Lex.
• Built face recognition solutions using Amazon Rekognition.

Telecom Customer Churn Prediction Project

Built a machine learning project to analyze telecom customer behavior and predict customer churn using Python. The project cleaned and transformed a dataset of over 7,000 customers, explored churn patterns through visualizations, and compared multiple classification models to identify which factors were most strongly linked to customers leaving.

Technically, I used Pandas and NumPy for data cleaning and preprocessing, converted categorical customer attributes into model-ready features with one-hot encoding, and used Matplotlib and Seaborn to visualize relationships between churn, tenure, contract type, monthly charges, and customer demographics. I then trained and evaluated several models, including Logistic Regression, Random Forest, Support Vector Machine, and an LSTM neural network, with the best model reaching around 82% accuracy.

Tech Stack: Python, Pandas, NumPy, Scikit-learn, TensorFlow/Keras, Matplotlib, Seaborn

Key Features

• Cleaned and preprocessed raw telecom customer data
• Handled missing values and converted categorical data into numerical features
• Visualized churn trends across contract type, tenure, charges, and demographics
• Trained and compared multiple ML models for churn prediction
• Interpreted model weights and feature importances to understand churn drivers