Praviveek Ray

Gemini Chatbot, Multimodal Assistant

• Built and deployed an assistant that handles text, images, and PDFs, using Transformers for language understanding and CNNs for image processing. • Ran it on Google Cloud and connected external APIs so routine customer support queries are answered without a human.

Bank Churn Prediction with Deep Learning

• Compared ANN, standard RNN, and LSTM on the same banking dataset to see whether sequence models pick up churn signal that a plain feedforward network misses. • Kept the preprocessing identical across all three so the difference came from the architecture alone, with LSTM performing best.

Customer Segmentation Model

• Grouped an unlabelled subscriber base into 6 behavioural segments, where GMM soft assignment picked up customers sitting between segments that K-Means forces into one. • Added a Random Forest churn classifier on top of the segments, with a dashboard for tuning the decision threshold.