May 2, 2018

Bonsai & Siemens: Autotuning CNC 30x Faster with Deep Reinforcement Learning

Artificial intelligence is transforming industries from manufacturing to energy and supply-chain, enabling subject matter experts within these fields to tackle their most challenging - and costly - business problems.

As highlighted in the announcement we made this morning, we partnered with the Siemens Digital Factory Motion Control team to work on one of these problems: combining Siemens subject matter expertise with deep reinforcement learning to calibrate a real CNC machine faster than an expert human operator.


Leveraging Bonsai’s Platform, including our unique Machine Teaching technology, we did just that.

A Siemens subject matter expert - an engineer with no background in AI - built an AI model that autotuned a real-life CNC machine more than 30x faster than the best operators. The machine was calibrated with expert-level precision, with one axis calibrating in just 13 seconds.

This achievement marks one of the first real-world industrial applications of deep reinforcement learning since DeepMind’s notable HVAC results in 2016.

If you happen to be at the O’Reilly AI Conference in New York make sure to come by my talk at 11am est; Deep reinforcement learning’s killer app: Intelligent control in real world systems.

Interested in learning more about the business impact of this accomplishment? Read the case study. You can also learn how to build deep reinforcement learning into your own real-world systems by visiting

Always. Be. Learning.

Stay up-to-date on our latest product news, AI industry highlights, and more!