Design and build the agent execution harness, ensure runtime reliability and observability at scale, build prompt/versioning systems, design evaluation frameworks, integrate frontier LLMs, and provide technical leadership for production agentic AI systems.