Explore recent advancements in Deep Reinforcement Learning and learn how to leverage TF-Agents, an open-source library for TensorFlow, in this 39-minute conference talk from Google I/O'19. Discover the clean, modular, and well-tested components of TF-Agents that can be mixed, matched, and extended to implement new RL algorithms. Gain insights into supervised learning, CartPole policy gradients, Atari Deep Q-Learning, and Minitaur Soft Actor Critic. Jump-start your Deep RL project with practical demonstrations and expert guidance from speakers Sergio Guadarrama and Eugene Brevdo.