Saturday October 30 9:00 AM – Saturday October 30 9:30 AM in Talks II

Large-Scale Production Reinforcement Learning with RLlib

Sven Mika

Prior knowledge:
No previous knowledge expected


We present RLlib, an open-source reinforcement learning (RL) library with a proven track record of having solved real-life large-scale industry decision making problems in production. We give two examples of how industry users utilized RLlib's different features and algorithms and thereby demonstrate how production RL represents the future of data engineering and data science.


In recent years, reinforcement learning (RL) has become a powerful item in our toolbox of machine learning methods. Its ability to produce end-to-end decision making solutions via learning by doing within a well defined problem environment makes it particularly attractive as an alternative to classic supervised learning methods. However, two issues remain problematic when it comes to using RL to solve real-world industry problems: 1) RL algorithms are hard to understand and therefore hypertune and 2) experiments need to run at scale in order to yield useful results within reasonable time. We will introduce RLlib (, an open-source RL library with a proven track record for solving real-life industry problems at scale. We will walk through two industrial RL case studies that leverage RLlib and share the lessons we learnt from accompanying them on this path. This talk is targeted towards data scientists and research engineers with some familiarity with machine learning concepts.