generative-ai system-design reliability llm-as-judge rules-engine state-machine ai-agents neuro-symbolic first-principles verification evaluation