For several months, USPS and NVIDIA engineering teams have developed AI models, which will help USPS read address labels faster and more efficiently. The system starts with high-performance servers powered by NVIDIA V100 Tensor Core GPUs and deep learning software, which trains multiple algorithms. The trained models are then distributed to NVIDIA EGX edge computer systems on nearly 200 USPS systems.
Delivery and testing of the system will begin this year, and AI is expected to be fully operational in the spring of 2020. While USPS will see an increase in efficiency, this is a big gain for NVIDIA as well. It shows that AI processes find more practical use.
The USPS is also experimenting with self-driving mail cars. This spring, it signed a contract with San Diego self-driving truck company TuSimple for a two-week pilot. This involved the company's trucks carrying USPS trailers on five round-trip trips between the postal service distribution centers of Phoenix, Arizona and Dallas, Texas.