You mentioned using async programming, cloud services, and caching to optimize API performance. Can you explain how you structured your FastAPI application to handle scalability and maintainability across features?

The FastAPI application is built around asynchronous endpoints using async def and the uvicorn event loop. We deploy the service on a cloud platform (e.g., AWS Lambda or GCP Cloud Run) behind a load balancer, which automatically scales based on request volume. Caching is handled with Redis for session data and response caching for idempotent endpoints. We use dependency injection to inject cache clients and database connections, keeping the code modular. For maintainability, each feature lives in its own module with a clear API contract, and we use Pydantic models for request/response validation. We also implement a shared middleware that logs request latency and errors, and we expose metrics to Prometheus. This structure allows new features to be added without affecting existing ones, and the async design ensures high throughput under load.