Lead the intelligence layer for a real-time voice AI agent: design and optimize model orchestration, prompt engineering and evals, build agentic workflow orchestration, ensure LLM reliability and regression testing, and select/optimize speech and LLM models balancing quality, latency, and cost.