Benji Barash

Robotics companies have to contend with a flood of log data that robots generate. Even a small number of deployments can quickly produce thousands of log files that contain complex, multi-modal sensor data. When managed incorrectly, this data quickly becomes a burden and can hide issues affecting system reliability and safety. Many companies are unable to answer simple questions like: “have we seen this problem before?

In this talk, we'll cover key lessons learned from productizing robots at Amazon, and show how you can leverage AI to search, transform and analyze robotics data at scale in the cloud. We’ll also highlight how getting this data under control can unlock the use of exciting new transformer models in robotics.