End-to-end MLOps pipeline for heart disease prediction using UCI Cleveland dataset with data validation, multi-strategy feature engineering, and hyperparameter optimization. Data flows from CSV through Great Expectations validation, then splits into three parallel pipelines: base features, PCA-reduced (5 components), and LDA-reduced (1 component), each training four models including Perceptron, Decision Tree, and Random Forest. Optuna optimizes hyperparameters with 5-fold cross-validation while MLflow tracks metrics, parameters, and artifacts for reproducibility. The best model registers in MLflow Model Registry and deploys via Streamlit app with drift detection, containerized in Docker, and hosted on AWS ECR with ECS Fargate for scalable inference. This architecture demonstrates production-grade MLOps best practices: automated validation, experiment tracking, model registry, and cloud-native deployment.