Programmatic generation of reinforcement-learning (RL) environments from real-world data for post-training evaluation. Targets ML/RL researchers and engineers needing reliable, data-driven evaluation environments.