Azura Fashion Group, a leader in sustainable and circular fashion technology, has announced the launch of a groundbreaking AI model poised to transform the fashion industry. This innovative technology aims to enhance sustainability and support circular economy practices, marking a significant step forward in the industry’s evolution.
The new AI model, integrated into Azura’s Transformer Platform, is designed to optimize the circular fashion supply chain. Utilizing cutting-edge AI technologies, Azura addresses the critical issue of poor product data, a major barrier for brands and marketplaces aiming to adopt sustainable practices in fashion.
Sam Wood, CEO of Azura Fashion Group, elaborated on how this innovative technology enhances sustainability and bolsters circular economy practices.
Also available in Spanish.
Vanessa Gatica: Azura Fashion Group, a leader in sustainable and circular fashion technology, has announced the launch of a groundbreaking AI model poised to transform the fashion industry. This innovative technology aims to enhance sustainability and support circular economy practices, marking a significant step forward in the industry’s evolution. The new AI model, integrated into Azura’s transformer platform, is designed to optimize the circular fashion supply chain. Utilizing cutting edge AI technologies, Asura addresses the critical issue of poor product data, a major barrier for brands and marketplaces aiming to adopt sustainable practices in fashion. Sam Wood, CEO of Azura Fashion Group, elaborated on how this innovative technology enhances sustainability and bolsters circular economy practices. What inspired Asura Fashion Group to develop this new AI model for the fashion industry?
Sam Wood: The biggest thing we we kind of, um, discovered when we were sending products to marketplaces and retailers was the inconsistency with the data. Um, we set this business up when we, uh, in 2019, off the back of some of the larger brands burning their excess stock, um, to kind of keep it out of discount stores. So we approached these businesses to allow them a different route by selling off season and past season products into new markets. Um, but how the products arrived at our doorstep or to our distributors doorstep or our wholesalers was it would come in in a pallet load and there would be very little data involved in those products. So for example, it would be an Armani t shirt, but all the information that would come with it would be Armani shirt. So when we were listing onto marketplaces, they required a four deep category level taxonomy. They required sleeve length, they required the model of the shirt, barcode and all this information that we didn’t actually have. So in the beginning we we were using different technologies and different, uh, softwares that were doing bits and pieces of it, um, but never really giving us the full picture, which was a bit of like a Frankenstein model. So we decided we needed to build our own system, and, um, we needed to tick all the boxes with the marketplace.
Vanessa Gatica: Can you explain how the AI model enhances sustainability and supports circular economy practices?
Sam Wood: So so how it does that is, again, we focus all our products are off season, past season or pre-loved. So how we do that is we build product information on products that the data might be lost or that we’re customers are even selling their bags back to us and they’ve lost the information of the product. So if you if you bought a Gucci bag and you sold it to Azura, usually you don’t have all the information with that, with that product description. So we use the AI model to then generate all that information. And then we use and we then use um, the Microsoft Azura portal to then send it out to marketplaces and retailers all over the world. So that product is getting a second life by selling through different marketplaces globally.
Vanessa Gatica: How does the AI driven product data transformation process work and what impact does it have on global marketplaces?
Sam Wood: So how it works is it would basically the product information would come into our system. So from a um, say a CSV or some kind of automated ingestion process, it then goes into an ETL kind of stage one where it gets mapped into our Asura Master catalog, um, using AWS Lambda functions. Um, we do a format normalization where we map it to our specific columns. So we say, all right, this is a brand name. This is its title. This is the taxonomy, the product description. It then goes into a ETL stage two process which is around um, the Asura data model. So an SQL database, um, where it does basic enrichment. So links to sizing charts, sizing, conversions. Like we get a lot of products that come in in Italian or in Italian sizing or in French language. So we convert that into a standardized model, which is say, English. And then we say, oh, this is small, medium, large, but it’s also an EU 42. It’s a Italian 54. Um, and we kind of build that model out. So it’s all around, uh, generating all those information. We then use AI to generate the titles, the descriptions, the bullet point summaries, the product composition, the formatting. We then use Google Translate to do the language translations. Um, and we do image enrichment through AWS as well. So that’s, that’s the ETL stage two process. And then we have a business modules which plug on to the Microsoft Azura platform, which is around the costing and pricing logic. So when products come into our system we have a landed cost in every single country. So we ship into around 30 countries around the world and we ship from around 28. So we have a different, uh, pricing model that goes through each entire process. So everything kind of comes in, um, and it gets a landed cost for, for each, each marketplace. And then we go into the marketplace mapping products and then it goes out. Being customers.
Speaker1: What challenges did Azura face during the development and integration of this AI technology?
Sam Wood: So the real the real challenges were, um, picking and choosing which which one was best, I guess. Um, each, each company, um, that we worked with had its pros and cons for different parts of their service. And rather than just going out there and building a Microsoft or just going out there and building on Google or Amazon, we decided to pull the best of both worlds in, um, all three worlds, actually, and pull every single model in. So we would use a Google image, we use Microsoft OpenAI, we use Microsoft Azure, we use Google AWS and Lambda functions. Uh, so we use the best of both worlds and building a system that Azure controls as well. So we’re no longer having to pay subscription costs to, um, software companies. It’s all around how we can really embed our own system and really go to market with our own platform, which is built from the ground up. So it gives the end customer a finished product.