Build next-generation, self-learning models using reinforcement learning techniques and best practices