Prime time

How Amazon's Houston fulfillment center uses AI technology and robotics to move millions of products

From robotics to artificial intelligence, here's how Amazon gets its products to Houstonians in record time. Photo by Natalie Harms/InnovationMap

Last summer, Amazon opened the doors to its North Houston distribution center — one of the company's 50 centers worldwide that uses automation and robotics to fulfill online orders.

The Pinto Business Park facility has millions of products in inventory across four floors. Products that are 25 pounds or less (nothing heavier is stocked at this location) pass through 20 miles of conveyor belts, 1,500 employees, and hundreds of robots.

The center also has daily tours open to the public. We recently visited to see for ourselves the process a product goes through at this Houston plant. From stowing to shipping, here's how packages go from your shopping cart to your front porch.

Starting with stowing 

Natalie Harms/InnovationMap

A product's first step in an Amazon facility is stowing. There's no categorization of the products — it's not like there's one floor for one type of item or anything.

"It's completely randomly stowed," says Donna Beadle, PR specialist for Amazon. "She could be stowing cat food on this floor, and so could somebody on floor two."

An Amazon employee would scan an item and stow it into an empty bin of her choosing — sort of. To prevent confusion, a light projected indicates bins that are off limits to stow the item. The light identifies bins that have similar products. Keeping similar products apart helps prevents mistakes for the employee who later pulls those items once its ordered.

The system also sees where the employee is putting each item, rather than having to scan each item and the bin as well. This is a newer feature — the facility originally opened with hand-held scanners.

"Our next generation workstation is that they don't have to hold that scanner — they have hands free," says Brenda Alford, regional communications manager at Amazon.

Robots on the move

Once the bins are fully stocked, the robot — which is the orange device on the bottom of the yellow bins — moves about the facility by scanning QR codes on the floor.

Should a product fall out, an employee wearing a special vest can enter to retrieve it. That vest will send off a signal to the robots, which will then decrease their speeds and come to a stop when the employee comes close.

"It's an extra measure of safety so that people can interact with the robots and feel safe," says Beadle.

Picking before packing

Natalie Harms/InnovationMap

Once an item is ordered, the bin with that item appears in the pick process at the center. The system tells the Amazon employee which item to grab and which bin to put it in. The bins will have products for multiple different orders — another employee later will separate it out later.

"Often we describe it as a symphony — our technology and our associates working together," Alford says, noting that sometimes the company might receive criticism about using robots over humans. "We can't do this without these humans.

Amazon employees receive their benefits from day one on the job, Beadle says, and they work four, 10-hour days a week.

"We feel like that way they have more time with their families — they get three days off versus two days off. And that gives them time to heal and rest up," she says.

Bin to bin and back again

Natalie Harms/InnovationMap

Once full, the Amazon associate will push the bin onto a series of conveyor belts. The whole facility has 20 miles of conveyor belts — much of which happens overhead.

The bins then zigzag toward the pack process, which is separated to different stations. There are single-product stations and multiple package stations. The system determines where the bin should go, and some stations pack products that are determined to need packing materials, while others do not.

Single-product packaging

Natalie Harms/InnovationMap

At the packing process, the Amazon employee is told which size box to assemble — he or she can grab a bigger box, but they can't select a smaller one. The tape dispenser doles out the correct size of tape for that box automatically.

Once packaged up, a sticker with a barcode is placed on the box. This code will later be used to print the label for shipping. At this point in the process, no personal information has been revealed to anyone. In fact, most packages leave the facility without any personal information being viewed by employees.

In an effort to reduce packing materials, some products are shipped in the container they came in. In that instance, the packer would just place the barcode sticker on the package before sending it on the conveyor belt.

"If we don't need another box for that product, we don't use one," Beadle says. "We work with companies to make that happen, so we don't have to use more boxes if we don't have to."

SLAM 


While the robotics aren't slamming labels on packages, the SLAM process (short for scan, label, apply and manifest) is the first step in the process that includes a customer's personal information. During this process, the barcode is scanned, the package is weighed, and the label is printed and affixed to the package using a puff of air.

A package might be automatically pulled from the line if something seems to be off in the package's weight.

"Say you bought toothpaste, and it says that toothpaste weighs 20 pounds, we know something's wrong," Beadle says. "Like maybe that it was a pack that didn't get separated."

If the package is kicked off, an Amazon associate, called a problem solver, will assess the situation and make it right before returning it to the conveyor belt.

Kicked into gear

Once labeled, all the packages are sent on their final conveyor belt ride. Using a scanning process, the packages are kicked by an automated foot that sends them into a line to be loaded into an Amazon truck.

If a package misses its chute the first time around, it makes the loop again. The system can tell if a package is caught in the loop for whatever reason, and a problem solver might be called to assess the situation.

Down the slide

Natalie Harms/InnovationMap

After being kicked off the belt, the package then slides down a spiral chute that, despite looking like a playground slide, is off limits to any humans wanting to keep their job.

"People ask if you can go down the slide, and we always say that on your last day of work," Beadle jokes.

On to the shipping process

Natalie Harms/InnovationMap

The packages leave the facility in Amazon trucks and head to one more pit stop before making it to the customer.

"They don't go directly to your house after this process," Beadle says. "They go to a sortation center."

This could mean a USPS or UPS stop, but it depends on where the customer lives.

Trending News

Building Houston

 
 

InformAI has three AI-based products geared at improving health care. Photo via Getty Images

In Houston, we’re lucky to have top-tier doctors in the Texas Medical Center, ready to treat us with the newest technology. But what about our family members who have to rely on rural hospitals? Thanks to one Houston company, doctors in smaller community hospitals may soon have new tools at their disposal that could improve outcomes for patients around the world.

Since InnovationMap last caught up with Jim Havelka, CEO of InformAI, two years ago, that hope has come far closer to a reality. InformAI is a VC-backed digital health company. Part of JLABS @ TMC innovation facilities, the company uses artificial intelligence to develop both diagnostic tools and clinical outcome predictors. And two of the company’s products will undergo FDA regulatory testing this year.

SinusAI, which helps to detect sinus-related diseases in CT scans, received its CE Mark — the European equivalent of FDA approval — last year and is being sold across the Atlantic today, says Havelka. He adds that in the United States alone, there are roughly 700,000 sinus surgeries that the product is positioned to support.

Another product, RadOnc-AI, is designed to help doctors prescribe radiation dose plans for head and neck cancers.

“Ideally the perfect plan would be to provide radiation to the tumor and nothing around it,” says Havelka. “We’ve built a product, RadOnc-AI, which autogenerates the dose treatment plan based on medical images of that patient.”

It can be an hours-long process for doctors to figure out the path and dose of radiation themselves, but the new product “can build that initial pass in about five minutes,” Havelka says.

That in itself is an exciting development, but because this technology was developed using the expertise of some of the world’s top oncologists, “the first pass plan is in line with what [patients would] get at tier-one institutions,” explains Havelka. This creates “tremendous equity” among patients who can afford to travel to major facilities and those that can’t.

To that end, RadOnc-AI was recently awarded a $1.55 million grant from the Cancer Prevention and Research Institute of Texas, or CPRIT, a state agency that funds cancer research. The Radiological Society of North America announced late last year that InformAI was named an Aunt Minnie Best of Radiology Finalist.

“It’s quite prestigious for our company,” says Havelka. Other recent laurels include InformAI being named one of the 10 most promising companies by the Texas Life Science Forum in November.

And InformAI is only gaining steam. A third product is earlier in its stage of development. TransplantAI will optimize donor organ and patient recipient matches.

“A lot of organs are harvested and discarded,” Havelka says.

His AI product has been trained on a million donor transplants to help determine who is the best recipient for an organ. It even takes urgency into account, based on a patient’s expected mortality within 90 days. The product is currently a fully functional prototype and will soon move through its initial regulatory clearances.

The company — currently backed by three VC funds, including DEFTA Partners, Delight Ventures, and Joyance Partners — is planning to do another seed round in Q2 of 2023.

“We’ve been able to get recognized for digital health products that can be taken to market globally,” says Havelka.

But what he says he’s most excited about is the social impact of his products. With more money raised, InformAI will be able to speed up development of additional products, including expanding the cancers that the company will be targeting. And with that, more and more patients will one day be treated with the highest level of care.

Trending News