AI is undergoing a paradigm shift, with breakthroughs achieved by systems
orchestrating multiple large language models (LLMs) and other complex
components. As a result, developing principled and automated optimization
methods for compound ai systems is one of the most important new cha