High-tech player in collections: Startup uses robots and AI to sort, analyze and sell trading cards

Matt Lubbers says the beginning of his new beginning was a visit to his friend Tim Clothier’s home, where the living room view of Mount Rainier was partially blocked. The problem? A mountain of trading cards from Clothier’s personal collection was on its way.

They weren’t just in the living room. The garage was full of cardboard boxes stacked on top of more boxes. A longtime collector, Clothier estimates his collection at about 7 million cards. Sorting and sorting them all by hand, he realized he could handle about 25,000 cards a week. He told his wife that it would take about 15 years to sort through them at that rate.

“I don’t think it was crazy to say, ‘What do you want here?’” Lubbers told GeekWire.

“My friends, when they’re done, I’ll be filtering and they’re running the other way,” said Clothier. “Matt was very curious and started asking questions, he said, ‘What do you think technology can do for you?'”

More than four years after that initial conversation, the startup’s founders discovered what the technology could do. Renton, Wash.-based Gradient is also active, using custom robots and artificial intelligence to help sort, analyze, list and sell sports trading cards, gaming cards, and more.

The goal is to capture a piece of the $15 billion US trading card market, help customers manage small and large collections, and easily and quickly get returns on eBay for sometimes forgotten treasures.

Card specialists and engineers

This hideout is located across the hall from the Seattle Sounders FC headquarters at the football club’s facilities in Renton – the Providence Swedish Performance Center & Clubhouse. Sounders majority owner Adrian Hanauer is an investor in Gradient, which raised $6 million from friends and family.

Clothier, the CEO, has known Hanauer since he was 15 years old. He spent 30 years at Pacific Coast Feather Co., the Hanauer family’s pillow and blanket manufacturing business.

Gradient’s expansive space looks like any office for an upstart company with a few notable exceptions. There are boxes upon boxes of trading cards everywhere, stacked next to rows of racks containing boxes of cards – 10 million in all and room to triple that.

A closer look at any open box or stack of cards reveals the faces of sports heroes past and present in baseball, football, basketball, hockey and more.

At a few tables there are staff shuffling cards by hand. Others at computer stations scan card files or write code that helps manage such a task. The environment is a mix of card experts and engineers.

And in one corner, there was the hum of eight robotic filters, comforting with a slight gust of wind and a breeze as parts moved cards back and forth in a custom-made machine that looked like something from a rock concert stage.

The system is the brainchild of Lubbers, a chief technology officer, computer visionary and AI expert who spent 15 years building complex systems and robots for autonomous vehicles and drones at ZF Group, Faraday Future, Voyage, Amazon Robotics and Zipline.

“We saw that there wasn’t much technology in this industry at that time. That’s what made us happy,” he said. “What if we could process cards very quickly? What if we could reduce the time it takes for a person, a customer or even a technician, to identify or price or write a card? That’s what we’ve built.”

Up to 100,000 cards per day can be processed by robots – and there is room to add more machines.

Lubbers is extremely protective of what he’s built, and wasn’t ready for GeekWire to capture photos or video of the robots at work.

Under bright lights, machines quickly move the cards through flatbed scanners to capture images of the backs of the cards as overhead cameras capture images of the fronts of the cards. Each card is cataloged physically and digitally.

While it may sound like fast-moving robots can be a recipe for disaster when mixed with delicate and sometimes important paper cards, the system is amazing. From the shape of the 3D printed trays where the cards are picked up and dropped, to the buttery soft suction fingers that gently lift each card, great care is taken to never mark or damage any card.

The collected images are instantly sent to a nearby server room where three custom supercomputers – using high-density configurations such as NVIDIA’s H100 or H200 chips – house six GPUs each. These machines handle all AI model training and hypothesis testing, going through 500,000 images per day to analyze and find cards against a database of 30 million variants.

Maintaining and managing the collection

Gradient joins a growing tech-heavy ecosystem where AI-powered platforms like Ludex, CollX, Card Boss and eBay’s inventory scan feature are already being used by collectors to quickly rate and value cards with quick scans through mobile apps. Gradient’s closest industrial competitor is probably TCG machines, which makes a robotic filter used by card shops to process thousands of cards per hour.

Gradient’s mission beyond showing how quickly and accurately it can process thousands of cards is to prove that it can store them efficiently, find them easily in QR-coded boxes and trays, and move them to the collectors’ market.

The company is just starting to attract customers, but the biggest one so far has given Gradient more than 500,000 cards to process.

Different subscription price tiers attract different customers and group sizes. Pay-as-you-go card scanning costs 40 cents per card. Premium level subscription is $9.99 per month for up to 10,000 cards; Pro is $29.99 per month for up to 30,000 cards; and Commercial is $99.99 per month for up to 100,000 cards. Levels include secure storage and other benefits.

Customers get access to a personal web site where they can manage their collections and see pictures of their cards, read card information, list on eBay through the Gradient Collects store, and monitor active listings and sales. The customer can choose one card or “send all my cards to eBay” and the Gradient system will issue such a request.

Gradient takes between 16% and 20% per sale, depending on the level of registration, with 13% or 14% of that covered by eBay.

The startup, which employs 25 people, broadcasts live auctions on eBay where auctioneers excitedly open packs of Pokémon cards from an indoor studio tucked behind a stack of boxes. And the company is building its own marketplace to offer customers the option of listing on Gradient, eBay or both.

Like a kid opening a new pack of cards at the corner mini mart, the possibilities with Gradient seem endless. Especially the kid, or, let’s face it, the old collector, who has finally uncovered those attic shoeboxes filled with thousands of cards and doesn’t know where to start.

“Our job is to help you digitize and value what you have, and then figure out what you want to do with it,” said Lubbers.