Implement a pipeline that automates the generation of reasoning-rich synthetic datasets using 'Chain-of-Thought' prompting for downstream model training.