Design a Secure Medical Audio Pipeline

You are reviewing a backend service for a healthcare application. The service accepts a medical audio recording, converts it to text, sends the transcript to an LLM API to generate a summary, and stores both the transcript and the summary for later retrieval. Assume the existing implementation is naive: a single request handler receives the upload, processes the entire file synchronously, calls external APIs directly, logs intermediate results, and writes all outputs straight into storage. Discuss the following: 1. What are the main security risks in this implementation, especially given that the data may contain highly sensitive medical information? 2. What are the main performance and reliability risks? 3. Design a production-ready system that is secure, scalable, and efficient. Explain the major components, data flow, storage choices, async processing model, access control, encryption, failure handling, observability, and how you would support large files and traffic spikes.