In this talk Sudeep will share an overview of the MLOps environment developed at TRI and discuss some of the key ways MLOps techniques must be adapted to meet the needs of high-stakes environments like robotics and autonomous vehicles. Environments like these add the challenges of edge or fleet deployment to those seen in traditional “online” environments, as well as the additional threat that if your model doesn’t perform well in production, people can get hurt. Attendees will learn how TRI has adapted MLOps techniques to achieve the desired level of efficiency while still ensuring model safety through a more rigorous approach to model testing, the use of open and closed loop checks, and the strategic use of a humans-in-the-loop for deployment to the fleet.
MLOps for High-Stakes Environments
Tuesday, January 19 |
10:40 AM -
11:10 AM PST