Large language model inference over confidential data using AWS Nitro Enclaves