Byte-Sized AI: Amazon Teams With Covariant on Robotics; Resonance Makes Made-to-Order More Accessible

Byte-Sized AI is a bi-weekly column that covers all things artificial intelligence —from startup funding, to newly inked partnerships, to just-launched, AI-powered capabilities from major retailers, software providers and supply chain players.

Amazon Sues NLRB After Board Determines Delivery Partners Are Joint Employers

Amazon gets cozy with Covariant

E-commerce giant Amazon announced on August 30 that it had minted a deal with robotics company Covariant , both on talent and on product, to make its robotics and automation capabilities stronger.

Covariant creates robotics foundation models, which allow full control over robots’ activities; the best ones allow control over multiple types of robots and different functions.

As part of the deal, Amazon will hire Pieter Abbeel, Covariant’s co-founder, president and chief scientist; Rocky Duan, co-founder and chief technology officer of the startup and Peter Chen, the robotics’ company’s co-founder and CEO. Amazon will also bring over about 25 percent of Covariant’s existing staff as part of its plan to grow its Bay Area robotics team.

Joseph Quinlivan, Vice President, Amazon fulfillment technologies & robotics, said he’s excited about the new talent and believes their expertise can take the giant even further.

“With some of the smartest minds, we will advance fundamental research, marrying our rich expertise to unlock new ways for AI and robots to assist our operations employees. Embedding Covariant’s AI technology into our existing robot fleet will make them more performant and create real world value for our customers,” Quinlivan said in a statement.

The company will also gain access to “a non-exclusive license to Covariant’s robotic foundation models,” according to its blog.

“Covariant’s models will help drive new ways to generalize how our robotic systems learn and provide dynamic opportunities for how we use automation to make our operations safer and better deliver for customers,” the company wrote in the blog.

Amazon already has over 750,000 robots working in its facilities and warehouses alongside human employees; by partnering with Covariant, the goalpost moves further in terms of what Amazon can build and do with robotics.

Resonance makes made-to-order more broadly available

Resonance , a technology and manufacturing company, announced Tuesday that it has publicly launched its AI-powered platform ONE, which allows fashion designers and companies to create items that are made-to-order upon purchase.

Brands can design their items directly in the platform, using either Resonance’s garments library or their own patterns. Designers can use ONE to add trims, change colors and otherwise alter how the garment looks. Once they’ve finalized the garment, it can be listed on their e-commerce site and made to order with each individual consumer purchase.

Lawrence Lenihan, the company’s chairman and co-founder, said that can help brands from both an inventory and an environmental perspective.

“Instead of forecasting , brands can use real customer demand to drive production, and never go out of stock on a product that could have driven more revenue. This entirely eliminates the inventory problem and frees up brands to be more consumer-oriented, creative, and truly sustainable ,” Lenihan told Sourcing Journal. “A brand could create capsule collections every week or offer limited-edition products based on what’s happening in culture—without ever producing a single garment until after it’s sold.”

Resonance already boasted customers like Rebecca Minkoff and Faith Connexion, but the newest announcement means the technology will become available to a broader swath of people in the fashion industry, whether major brands or individual designers just starting out.

According to the company’s website, access to ONE costs $100 each month, plus cost of goods sold (COGS) and a 10 percent revenue share.

Lenihan said he and his co-founder, Christian Gheorge, have spent over seven years working on the technology behind the system, which has a foundation of artificial specific intelligence.

This Tool Uses Gen AI to Link Separate Instances of Organized Retail Crime

“ONE has been learning from a trove of proprietary data in order to minimize waste in the industry at every step. It’s time to open this system up to the world and enable more brands to operate without inventory, minimums or waste,” he told Sourcing Journal.

Google expands virtual try-on tool to include dresses

Google has updated its virtual try-on tool to allow users to see how dresses fit generative AI -powered models of varying sizes.

The technology giant already offered the capability for other clothing items, particularly men’s and women’s tops, but other items, like dresses, can be more difficult to show in realistic fashion with generative AI.

Google said dresses proposed two key challenges: dresses often have more details than standard tops or blouses, and dresses cover more of the human body, which means they can be more difficult to generate images of without tainting or losing the human-like features of the AI models.

Those hurdles are likely why it took Google some extra time to develop virtual try-on for dresses. It uses a technique called diffusion, which generates individual pixels from the bottom up, to create the image that end users see.

To overcome the issues it had with keeping models’ identities intact, Google developed a new system, which it calls the VTO-UNet Diffusion Transformer (VTO-UDiT).

“[This technique] isolates and preserves a person’s important features. So while we train the model with “identity loss” in place, VTO-UDiT also gives us a virtual ‘stencil,’ allowing us to re-train the model on only the person, preserving the person’s face and body. This gives us a much more accurate portrayal of not only the dress but just as important, the person wearing it,” Ira Kemelmacher-Shlizerman, principal scientist, shopping, and Lilian Rincon, vice president, product management, wrote in a blog.

When shoppers search for dresses on the Google Shopping tab, they are presented with some options that offer try-on; if they choose to use the function, it shows them models with different skin tones and body shapes to select from between sizes XXS and XXXL.

According to Google, some of the brands already onboarded include Simkhai, Boden, Staud, Sandro and Maje.

Startup scores $8.2 million for ‘AI concierge’

Rep AI announced that it had secured $8.2 million in funding in an undisclosed round. Osage Venture Partners led the round, with support from Orzyn Capital and Flashpoint Capital Ventures.

The startup has built a tool to aid brands and retailers with answering customers’ questions and needs during e-commerce sessions, and the funding will head toward building that tool out further, the company said.

According to Rep AI, already the technology has been able to, on average, increase retailers’ conversions up by 22 percent. Yoav Oz, co-founder and CEO of Rep AI, said the technology will continue to drive outcomes for brands and retailers looking to connect more closely with their customers.

“While technology has changed immensely in the past 15 years, e-commerce has largely stayed the same—and so has its historically poor conversion rates. Rep AI is the evolution the industry has so desperately needed,” Oz said in a statement. “Unobtrusive and completely customizable to match each merchant’s product catalog and brand personality, it brings any sized business the benefits of a highly skilled sales expert that also happens to be infinitely scalable.”

Rep AI combines large language model ( LLM ) technology with behavioral AI to understand how a shopper browses and discovers products, in turn allowing the technology to identify touchpoints where the consumer may want assistance or recommendations; from there, it offers to help the consumer throughout the rest of the browsing session.

“Rep AI fills the need to consult with a professional while purchasing online. Especially for first-time shoppers at any given site, consumers want more information and assurance before making a buying decision,” said Noam Wolf, partner at Flashpoint Venture Capital. “In that perspective, e-commerce today is broken, leading shoppers to experience distrust and anxiety. Rep AI solves this problem, assisting both shoppers and e-commerce businesses to have a better experience.”

Walmart Enables Marketplace Sellers to Fulfill From Third-Party Sites