System Design · Amazon · Hard
Design a comprehensive quality assurance platform for large language models in a team that trains and serves these models. The platform must automatically and systematically evaluate correctness, safety, and regression risk for newly trained model versions and for responses generated in production. Your design should cover the following areas: 1. Validation scopes and layers Define what should be checked at different abstraction levels. At minimum, address: low-level…
Checking your access…