Deploying Secure Generative AI Apps

This unit covers end-to-end implementation and deployment practices for generative AI applications using Amazon Bedrock playgrounds and a Flask web app. Students will automate continuous learning and security/privacy monitoring, experiment with prompts and model configurations, and build a working movie-recognizer prototype while applying iterative evaluation and mitigation strategies.

Learning Objectives

• Analyze potential security, privacy, and compliance risks in LLM deployments and design an automated continuous-learning and monitoring workflow to detect and mitigate those risks.
• Apply AWS Bedrock playground tools and experimentation workflows to prototype, test, and document prompt variations, model parameters, and performance metrics.
• Demonstrate the development, local testing, and basic deployment of a Flask-based generative AI web application (movie recognizer) integrating inference, prompt handling, input validation, and simple privacy safeguards.
• Evaluate model outputs and prompt iterations to improve accuracy, reduce hallucinations, and enforce privacy/compliance constraints through feedback loops and automated monitoring.