Valmont
Herd Sense Story
Heifer #445 wanders through its pen in a feedlot in Yankton, South Dakota, unaware that it is being constantly assessed by a computer perched on a nearby pole.
The computer analyzes video taken by its camera to determine that #445 weighs 752 pounds. It keeps track of how much #445 is eating and drinking. And it watches #445’s posture and gait for telltale signs of illness or injury.
“It’s watching all the animals every day, all day,” says Trevor Mecham, a Valmont Industries vice president who led the creation of the monitoring system.
Valmont’s Herd Sense system is a remarkable use of artificial intelligence and facial recognition that could result in healthier cattle, more efficient operations and better returns for ranches and feedlots.
But it wouldn’t have been possible without Valmont obtaining a major boost in computing power — allowing developers to make continuous updates to the model that scans and analyzes cattle health.
To get that GPU power, the Fortune 1000 company chose Scott Data Center and its innovative colocation-based AI services.
Scott Data has provided data services to companies, organizations and the U.S. military for nearly two decades. It operates an 110,000-square-foot data center in Omaha that is certified as Tier III by the Uptime Institute for its reliability and security.
Recently, as the demand for AI computing has grown, Scott Data has been adding cabinets full of high-speed AI processors. The AI chips have enormous requirements for power and cooling, but the data center has the ability to handle those demands.
Customers like Valmont can essentially rent a slice of those processors to fulfill their artificial intelligence needs.
Scott Data also provides comprehensive service to help both large and small companies navigate the process of adopting AI.
Industry expert Bill Kleyman says Scott Data’s “GPU as a service” slogan highlights its uniquely valuable niche among data centers. It comes at a time when AI is bringing revolutionary changes and companies like Valmont are trying to take advantage of the opportunities.
“This is not a technology shift,” Kleyman says. “It’s a fundamental shift in humanity’s way to interact with data. We’re just kind of scratching the surface of it.”
Artificial intelligence lets businesses come up with new useful products or more efficient processes, growing their companies and increasing profits, he said.
Yet it can cost millions for a business to set up its own stand-alone AI system. And Kleyman says not all data centers are equipped to add AI capacity as Scott Data is doing.
Some traditional data centers, built to house racks of standard computer servers, can’t afford to add costly AI graphic processing units. Nor can they handle their considerable power and cooling needs.
But Scott Data has invested millions in new GPUs that fit into its already-robust infrastructure. The Omaha center even innovated an industry-leading approach to maximize the density of high-speed GPUs.
Other data centers may have the supporting infrastructure and high-speed chips, Kleyman note, but they aren’t willing to let a customer bring in its own physical equipment such as firewalls or data storage devices. That can be a problem for businesses that must comply with strict security regulations. Scott Data accepts other equipment on the data floor and can segment it off as needed.
And some data centers certainly have the capacity to provide AI computing, but they’re not set up to offer “white-glove” managed services to customers, including those who are new to artificial intelligence.
That’s where Scott Data really shines, Kleyman says.
“It’s tremendous because you’re talking about businesses that haven’t been in AI,” Kleyman says. “They feel like they need to get into, but they have a huge upfront (capital) cost if they’re going to do it themselves. They may not have the expertise, and they don’t have the infrastructure to surround all that.”
Kleyman says Scott Data helps such AI newcomers “get their toe wet” with the technology. Customers can get technical guidance in building an AI application, and they pay only for the amount of computing they need now — with the ability to increase that as time goes on.
“We can meet them where they are,” says Dominic Romeo, Scott Data’s director of technology services.
Says Kleyman:
“No one else is offering a service like this.”
Mecham agrees. He says Valmont looked at its options for AI computing and couldn’t find any other data center that would meet the company’s needs for the new Herd Sense product. And despite Valmont’s size and internal IT expertise, it didn’t make sense for the company to create its own AI operation either.
“That’s not what our traditional model is,” he says. “Why would I start at the starting line when I could start at the finish line, utilizing somebody like Scott Data Center, where their core competency is exactly what I need.”
What Valmont needed was faster AI processors so that its Herd Sense developers could turn Mecham’s idea into reality.
Mecham heads Valmont’s innovation group and was looking for new uses for Valmont’s products, which include galvanized poles for streetlights, utilities and transportation as well as irrigation pivot systems.
What if one of those poles, Mecham wondered, could be placed in the corner of a feedlot to do a whole range of monitoring of animal biological behavior?
Mecham had grown up on a farm in southern Idaho, raising cattle and potatoes. It taught him a good work ethic, but he didn’t enjoy all the laborious tasks. And as an adult working at Valmont, he thought the task of caring for a cattle herd might be improved through automation.
“How long do you think it would take me to monitor each one of those animals in a pen as a pen rider, as a cowboy going in there?” he asks.
“What am I really going to be [able to] identify and say? ‘I think that one’s sick, but I don’t know for sure. It looks like it’s limping, but I’m not exactly sure.’ And I’ve got to do that through all of this feedlot operation.”
A single pen rider has limited time to assess each animal during a shift, he says. But continuous video monitoring can watch them all day, regardless of weather conditions and other duties.
Valmont’s system on the feedlot pole is about the size of an Xbox gaming system. It analyzes the video of the animals below and sends results to the rancher or feedlot operator.
If the camera spots a possible problem with an animal — a limp, for example, that looks like it might be hoof rot — the system keeps its eye on it. If it persists, Herd Sense can alert a pen rider to take action. If the “problem” turns out to be just an anomaly, system avoids sending out a false positive.
Providing that sort of curated information makes a pen rider more valuable, because they can focus on specific issues that affect the health of the herd and profitability of the feedlot or ranch.
“It’s making sure that that employee’s time is being spent in the most effective way,” Mecham says.
To get to that point, Valmont’s experts came up with a unique approach to building a computer model that could recognize an ailing cow or make other assessments of each animal.
Instead of starting from scratch to create that model, the Valmont team used actual feedlot videos to create digitized versions of healthy cattle and unhealthy ones. Then it trained the computer model to tell the difference.
The testing also involved using a scale next to a water trough to compare cattle weights estimated by the model with the actual weight of the animal. If they didn’t match, developers knew they had to tweak the algorithm so its estimates would be more precise.
Doing so required massive amounts of machine learning. Before Valmont moved its AI work to Scott Data, it was using slower processors that had to run for weeks in order to incorporate and test each addition or revision to the model.
After the shift to Scott Data and its faster GPUs in March 2024, those revisions only take a few hours.
“For my technology teams, life changed forever for the better,” Mecham says.
“There’s no way I could do it without this kind of computation.”
He also describes the process of shifting the company’s Herd Sense models to Scott Data’s AI processors as very smooth, thanks to the data center’s managed services staff.
“They’re really providing me a turnkey solution,” he says. “It just optimized, because this is what they do for a living.”
All those AI-fueled updates to the Herd Sense model make the system more accurate and more predictive, and allow developers to address customer requests for additional information or analysis that accounts for different breeds.
Once updated, Valmont then uploads the improved software via satellite link to each box on a feedlot pole, where it goes to work immediately to monitor cattle.
Herd Sense has been under development at a handful of test sites but is about to be released nationally.
Mecham says Valmont’s underlying technology and monitoring system also has a lot of potential uses besides livestock management.
The approach could be adapted for quality control in a manufacturing facility, for example. Or it could help cities with dynamic traffic management that responds quickly to congestion at an intersection.
Mecham likes that Scott Data’s approach to providing GPU service will allow Valmont to add more computing power as those new uses are developed.
“I’m just going to need them more and more every year,” Mecham says.
Author Name
Author bio goes here. Click edit button to change this text. Lorem ipsum dolor sit amet, consectetur adipiscing elit. Ut elit tellus, luctus nec ullamcorper mattis, pulvinar dapibus leo.
