CX Cortalyst: AI-Powered Customer Review Triage & GenAI Response System | AAI-540 Final Project

CX Cortalyst is an end-to-end MLOps system built on AWS SageMaker that automatically classifies customer reviews as Positive or Negative and drafts a personalized GenAI response for every negative review — ready for human agent review. This video demonstrates the full production pipeline: Business problem & system overview Architecture walkthrough SageMaker Feature Store Model Registry & evaluation metrics Batch Transform predictions CloudWatch monitoring dashboard Drift simulation & alarm validation CI/CD pipeline: success & failure states GenAI response generation (FLAN-T5-XL via JumpStart) Future improvements & risks Wrap-up Built with: Amazon SageMaker (Training, Processing, Batch Transform, Pipelines, JumpStart, Feature Store, Model Registry, Model Monitor), AWS Glue, Amazon Athena, Amazon CloudWatch, XGBoost, sentence-transformers (all-MiniLM-L6-v2), and FLAN-T5-XL. Dataset: Yelp Open Dataset (review, business, user, tip, checkin) GitHub repository: https://github.com/jagadeesh-usd/aai540-se... Project for AAI-540: Machine Learning Operations, University of San Diego. Author: Jagadeesh Kumar Sellappan | Team/Group #8