Robotics models need very fast inference, and on-premises or on-device serving is the fastest way to achieve it.

“our model is pretty smart so that's okay but uh in general like if you think about very dynamic tasks like for example you want your robot to like you know hold a i don't know a tennis racket and you know juggle the ball right it's very dynamic very reactive you can't just like the physics gravity doesn't wait for you right so for that reason robotics model you”

“though it needs to be able to inference very fast and the fastest way to do that is by serving on-prem like on device right because if you call a model by”