Sparsha Srinath

Credit Card Defaulters

• Achieved 85% accuracy in credit card payment default prediction, managing 32,561 instances with robust validation and preprocessing, utilizing KMeans clustering for model enhancement.
• Implemented a database-driven pipeline for 99% data accuracy, handling missing values and numeric scaling, while training Naïve Bayes and XGBoost models for clusters with an average AUC score of 0.87.
• Orchestrated deployment on Heroku cloud with a Flask web app, showcasing Git and Heroku expertise for continuous model improvement and management.

Chatbot for Course Information

• Developed an Amazon Lex-driven chatbot, adeptly collecting 100+ user profiles while ensuring their privacy and data security.
• Orchestrated an automated email system that effectively disseminated personalized course details, garnering highly positive user reviews.
• Skillfully harnessed AWS Lambda and related services to facilitate seamless interactions with users, and seamlessly integrated the chatbot with Facebook Messenger, significantly amplifying user engagement and efficiency in addressing inquiries.

 Cafeteria System

• Developed an online cafeteria system using Django, integrated Twilio for order notifications, Mailchimp for customer engagement, and Stripe for secure payments.
• Employed Python and Django for a robust backend, ensuring 99.9% uptime and 95% customer satisfaction, with deployment on Heroku.

Microsoft Certified: Azure Data Fundamentals

Singh, A., Srinath, S., Arasu, V., & Thomas, N. K. (2022, July). Machine Learning on Web: Skin Lesion Classification using CNN. In 2022 International Conference on Inventive Computation Technologies (ICICT) (pp. 260-265). IEEE