USPS
The United States Postal Service, or USPS, is relying on artificial intelligence-powered by Nvidia’s EGX systems to track more than 100 million pieces of mail a day that goes through its network. The world’s busiest postal service system is relying on GPU-accelerated A.I. systems to help solve the challenges of locating lost or missing packages and mail. Essentially, the USPS turned to A.I. to help it locate a “needle in a haystack.”
To solve that challenge, USPS engineers created an edge A.I. system of servers that can scan and locate mail. They created algorithms for the system that were trained on 13 Nvidia DGX systems located at USPS data centers. Nvidia’s DGX A100 systems, for reference, pack in five petaflops of compute power and cost just under $200,000. It is based on the same Ampere architecture found on Nvidia’s consumer GeForce RTX 3000 series GPUs.
The algorithms are then deployed and used on a network consisting of 195 distributed Apollo servers created by Hewlett Packard Enterprise, with each server equipped with four Nvidia V100 GPUs. The result is a system called the Edge Computing Infrastructure Program, or ECIP, which tracks items for the postal service.
Combined with optical character recognition, these systems can locate missing mail that goes through the USPS network. In the past, it would take days to track down packages, but A.I.-powered technologies is reducing the manhunt for lost packages down to just mere hours.
“It used to take eight or 10 people several days to track down items, now it takes one or two people a couple hours,” said Todd Schimmel, the manager who oversees USPS systems in an Nvidia blog post.
The use of graphics-accelerated computing was perfect for this A.I. task, as it would have taken a network of 800 CPUs more than two weeks to complete to do the same thing that four Nvidia V100 Tensor Core GPUs could accomplish in a span of 20 minutes on the HPE Apollo 6500 server.
Each of the Apollo edge servers process 20 terabytes of images daily from more than 1,000 mail processing machines, Nvidia stated. The USPS contract was awarded in September 2019, Schimmel said, and the hardware installation was completed by August 2020.
“The app that checks for mail items alone requires coordinating the work of more than a half dozen deep-learning models, each checking for specific features,” Nvidia stated. “And operators expect to enhance the app with more models enabling more features in the future.”
In addition to mail tracking, the USPS is working on different use cases for A.I. at the edge, ranging from enterprise analytics to finance and marketing. There are currently 30 applications planned for ECIP, and the postal service hopes to implement a few of its ideas this year. One such idea would have A.I. analyze if a package has the correct postage for its size, weight, and destination, and another ECIP app could decipher a damaged barcode, for example.