As Shopin approaches its official launch, the company added aspects of its core proprietary technology infrastructure to GitHub this week as part of a third-party audit performed by a representative from Novoa Media.
The audit, while mostly internal, included aspects of Shopin’s Artificial Intelligence, personalization algorithms, and back-end software. Over time, additional code will be shared publicly on GitHub as examples of the significant product development milestones Shopin is spearheading for the technology and crypto communities.
As with many companies in the digital space, most projects are based on open source software that is then enhanced to fit that company’s particular IP. Shopin is committed to supporting the open source community, but the Shopin business model, while built on top of open source software, is not open source, and much of company’s code cannot be shared without disclosing proprietary company information.
“We’re excited to share some of what we’ve been working on at Shopin as we develop projects using blockchain technology to enhance the shopper and retailer experience and keep shopper data private,” said Jeremy Harkness, Blockchain Technology Officer for Shopin. “Our GitHub presence is important because it’s a repository for all our code and a great deal of our company value is stored there. However, since the majority of our work is proprietary, we cannot show projects in their entirety, and we must be somewhat protective of what we display. That said, we will make every effort going forward to share insights when appropriate to keep our community informed of continued innovations underlying the Shopin vision.”
In addition to receiving accolades from the auditor for building out a robust A.I. product — at a time when many companies lay claim to such work but very few fulfill on it — the audit included a thorough analysis of Shopin’s unique solutions and how they are transforming the retailer and shopper experience.
According to Shopin’s development team, in order to personalize recommendations for shoppers, it is vital to understand the semantics of fashion apparel. Shopin developed a suite of deep learning algorithms that build a knowledge base of fashion compatibility by mining fine-grained details from a large, curated collection of products.
The Shopin Github audit analysis went into further depth on Shopin’s shopper profile and recommendation engine and its ability to accurately predict attributes and retrieve products and cross-category products from retailers and websites based on nuanced details, including patterns, color variations, the shape of a neckline on a dress, the length of sleeves, and other precise details that support relevant product recommendations for fashion-minded users.
According to Divakar Rayapaty, CTO and CoFounder of Shopin, one of Shopin’s most challenging endeavors in product development has been to computationally model the thinking and decision-making process of fashion stylists.
“Our recommendation engine enables personalization of the highest value in terms of relating to the specific user’s needs and interests,” said Rayapaty. “If they are reluctant to show off their arms, for example, then we cannot recommend a product that is sleeveless. The needs of this industry are extremely nuanced in terms of style preferences and exacting details, all the way to the fabric quality itself. So we’re not just capturing stylistic attributes, we’re capturing personality attributes that play into a buying decision.”
Shopin’s CEO and Founder, Eran Eyal, pointed to some of the challenges retailer data presents and why visual A.I. plays such a critical role.
“Humans are both creative and inconsistent in the way that we observe and describe our visual experiences,” said Eyal. “We tend to add creative flair and there is no uniform language across retailers and brands. But what is true, is that a picture is worth a thousand words. Our visual A.I. constructs 25–36 attributes per product that are normalized from the outset. For true matching to happen between a user and product, visual A.I. is an absolute necessity. Where does it go from here? Well, once we have mapped all of fashion and then home goods, all content across publishers can become identifiable and even purchasable with little effort. Essentially we are creating a consistent universal language for the new, distributed internet. We’re not just building the groundwork for a distributed Amazon of physical content (products), but for virtual content as well.”
To stay in touch with Shopin and future product developments, please reference our website, blog posts, and please join the active conversation on our Telegram group or Telegram channel, Twitter, Facebook, LinkedIn, Reddit, Steemit, and YouTube.