Bengaluru, Karnataka, INDIA: For somebody pushing 40, stepping right into a health club for the primary time is usually a nerve-wracking expertise.
As this author realized, even earlier than self-doubts crept in about their probabilities of surviving an hour on the health club, the bigger looming query was what to put on and never look misplaced.
With completely no thought what they had been in search of, this author turned to an AI purchasing assistant by Myntra, India’s largest on-line trend retailer, and typed, “I’m looking for clothes I can wear to work out in the gym.”
Surprisingly, the AI assistant understood precisely what this author wanted and got here up with jerseys that might wick off sweat, compression t-shirts, self-proclaimed snug trackpants that wouldn’t prohibit motion, sneakers that might make you run higher, health bands and all kinds of drugs a beginner couldn’t have imagined they needed or wanted.
With the purchasing cart full and the pockets considerably empty, this author was prepared for a brand new starting.
What the AI assistant did – convert an summary consumer question into actionable outcomes – is sport altering for the style business. Conventional search works finest with particular key phrases – a blue t-shirt from a specific model, say.
It goes a couple of steps past typical search. It makes use of generative AI to reply to extra open-ended questions like what to put on for a specific competition or a cricket match and even the trending trend in a metropolis.
“This is big,” stated Arit Mondal, director of product administration at Myntra, “Why? Because, this is the first time we have a solution, which is solving the unsolved ‘search’ problem in the fashion, beauty and lifestyle industry. And it’s live for customers at scale.”
![Portrait photo of a man standing in front of mannequins](https://news.microsoft.com/source/asia/wp-content/uploads/2023/12/Myntra-2-1024x683.jpg)
Since the start of on-line trend retail, trying to find merchandise has been very comparable to looking for some other piece of knowledge on-line. You attempt a set of key phrases and hold refining your search with totally different key phrases and preset filters.
A seek for a branded, blue t-shirt works effectively as a result of the key phrases are already a part of the product catalog.
But that’s not at all times how individuals store in the actual world. Some consumers solely have a obscure thought what they need – as an example, garments for an upcoming trip or a rock live performance.
The typical technique of looking by key phrases fails spectacularly on the subject of the second type of buyer because the search strings they use are usually not retrievable straight from the knowledge saved within the product catalog.
Until now.
When generative AI – constructed on massive language fashions (LLMs) that synthesize huge troves of information to generate, textual content, photographs and extra – first made information final yr, the workforce at Myntra shortly started fascinated about how they might leverage it to reinforce buyer experiences.
When Myntra organized a hackathon in February this yr, a gaggle of engineers from the corporate’s search workforce determined to make use of Azure OpenAI Service to resolve the summary search drawback and unshackle customers from the cuffs of key phrases.
![Photo of a group of people looking at a computer](https://news.microsoft.com/source/asia/wp-content/uploads/2023/12/Myntra-3.jpg)
They had been pleasantly shocked to see how ChatGPT, the generative AI service obtainable by means of Azure OpenAI Service, may synthesize pure language prompts. They requested ChatGPT in regards to the look of an actor from a latest film and it may inform it consisted of a bomber jacket, gloves and aviator sun shades.
“And this is the information that Myntra’s existing catalog didn’t have,” stated Swapnil Chaudhari, an engineering supervisor at Myntra.
Over two days, his workforce took over a convention room and saved attempting new prompts – textual content that generative AI may perceive – to see what outcomes they obtained. This was new territory – they usually didn’t know the way far they might push.
“We were surprised to see the results. It was able to answer questions like clothes to wear for regional festivals like Pongal and Onam,” stated Pragna Kanchana, a frontend engineer at Myntra.
On a whim, she tried to go looking in Hindi with sardiyon ke kapde, which in English interprets into winter garments. And it understood it!
The workforce then obtained entry to Azure OpenAI Service’s playground that allow them do rather more than was doable with ChatGPT alone.
“Leveraging Azure OpenAI Service, we were able to plug in different large language models in the same prompt and figure out which model worked best for our use case. So, we had a lot of freedom to compare and choose the right model,” defined Santanu Kanchada, a backend engineer within the search engineering workforce.
The workforce knew they had been on to one thing huge. They wrote the code in a day, and inside two days they’d a working prototype of a brand new characteristic that enabled customers to go looking with pure language.
“If it were not for GPT models, we’d have to first retrain the model using Myntra’s catalog and then wait and check the results with our expectations. But the pre-trained models already available with Azure OpenAI Service were already performing quite well,” added Chaudhari.
Over the subsequent 5 weeks, a number of groups throughout engineering and product growth fine-tuned each the backend and the consumer interface for the AI purchasing assistant.
“Myntra’s systems are on Azure and deploying Azure OpenAI Service was as seamless as deploying another server and it gave us a secure way of using generative AI,” defined Vindhya Priya Shanmugam, director of engineering at Myntra.
![Portrait photo of a woman standing in front of a wall of oversized colorful spools of thread](https://news.microsoft.com/source/asia/wp-content/uploads/2023/12/Myntra-1.jpg)
Post the hackathon, the search engineering workforce saved refining the prompts to get helpful outcomes for customers. One of the issues, as an example, was how to make sure that the response to a consumer’s question resulted in garments for less than the gender the consumer is in search of.
In the weeks resulting in the launch, they skilled the system on Myntra’s catalog and added guardrails so the outcomes had been restricted to the catalog.
The AI purchasing assistant was launched on the Myntra app in late May, simply in time for considered one of their largest marquee occasions, End of Reason Sale (EORS). It included pattern prompts that gave customers an thought of how they might use conversational language fairly than key phrases.
Since then, Myntra has already seen search queries broaden, providing new alternatives for product discovery. For occasion, when somebody searches for garments they’ll put on to a seashore, not solely seashore put on but in addition equipment like hats, sun shades and footwear pop-up.
It has been phenomenal for Myntra.
“Users who shop using the AI shopping assistant are three times more likely to end up making a purchase,” stated Mondal. “Because it also helps users discover a complete look from multiple categories of products, we’re seeing that on average they add products from 16 percent more categories than usual.”
While this author’s health transformation journey continues to be questionable, a number of groups at Myntra are already constructing new options based mostly on generative AI.
One of them will permit customers to decide on totally different classes of merchandise – tops, bottoms and equipment, for instance – and see how they give the impression of being collectively in an outfit. Myntra plans to additional improve it by introducing voice search and supply customized outcomes. They are additionally taking a look at how they’ll use generative AI to assist the client help groups.
Top picture: Myntra’s AI purchasing assistant powered by Azure OpenAI Service lets consumers uncover an entire look utilizing pure language prompts that may embody locations, festivals, or different events. Photo by Selvaprakash Lakshmanan for Microsoft.