How to train and scale AI math/coding agents using VeRL on any AI infra

VeRL and SkyPilot work together to train AI agents that solve math problems and write code using reinforcement learning, with the entire process automated through a single command that finds GPUs, prepares data, trains models, and serves them across any cloud infrastructure. The tutorial demonstrates building both a math agent that provides detailed step-by-step solutions and a coding agent that generates executable code, showing how trained models outperform base models by offering more structured reasoning and practical solutions.