Mapping the mind

Would you trust a machine to be your shopping partner? Can it tell a shawl lapel from a peak lapel for your jacket? Is it capable of suggesting a pair of shoes to complement your trousers? Mad Street Den (MSD), an Indian-American artificial intelligence (AI) company, says it has a machine that can do all that—and more.

To be precise, Vue.ai is not a machine in the conventional sense. What the company’s founders, the husband-and-wife team of Anand Chandrasekaran and Ashwini Asokan, have actually developed is a neural network—a computational approach modelled on the brain—that spots trends and patterns to help retailers improve a customer’s shopping experience. In other words, with Vue’s brain in place, your retailer can track what you select, what you don’t, your colour preferences, and more—both on the web and in brick-and-mortar stores.

“Vue.ai is like the friend who goes shopping with you and understands what you like, recommends products that you’d like, as well as curates ensembles that would look great on you,” says Yashika Punjabee, co-founder of Indian online fashion and lifestyle retailer TheLabelLife.com, which uses Vue.ai’s technology to let its style editors like Malaika Arora Khan guide buyers.

So, how does Vue.ai work? On the web, it is fairly predictable. Vue.ai’s visual search and emotion-recognition capabilities allow it to record your online activity and shopping history, recognise your taste, and suggest clothes accordingly. In a store, it gets a little too close for comfort: cameras and sensors watch your actions such as what section you walk up to, or what colour catches your eye. Stores can use this data to decide their display and design.

Vue.ai was launched just last year, but it already has a string of retailers in the bag. They include U.S. seller of secondhand clothes thredUP; Britain-based Villoid, a fashion app co-founded by supermodel Alexa Chung; online e-commerce platforms Wadi and Namshi in West Asia; furniture and decor seller HipVan in Singapore; and Indian online platforms Tata CLiQ and The LabelLife. “We are also working with a large [brick-and-mortar] retailer in the U.S., and hope to make it public mid-2017,” says Asokan, the company’s CEO.

That’s a long list for a startup that Asokan and Chandrasekaran set up barely four years ago with offices in Bengaluru and Chennai. It raised seed capital of $1.5 million (Rs 9.5 crore) from Exfinity Venture Partners in 2015, and an undisclosed amount in venture funding from investors including Sequoia Capital and Exfinity in 2016. Since then, the company’s headquarters have moved to San Francisco; it also has an office in London.

For years, AI was largely an academic subject, which only conjured up images of humanoid robots to a larger audience. Today, it is no longer an abstract idea in the realm of science fiction, but a fast-developing field in the world of business. Global giants such as Google and Apple have AI to develop a host of products from virtual assistants such as Google Assistant and Siri to driverless cars.

The retail business is one of the biggest customers of AI-powered tools. Although India has dozens of software companies that have developed AI-based applications for sectors from health care to banking, MSD is one of the few to realise the potential of retail. According to a report by the Federation of Indian Chambers of Commerce and Industry and consultancy firm PwC, India’s retail market is expected to reach $1 trillion by 2020, and the e-commerce market is expected to reach $125 billion in terms of gross merchandise value.

“Retailers have so far invested aggressively in analytics to provide personalised experiences to consumers,” says Manish Bahl, senior director, The Center for the Future of Work at IT services provider Cognizant Technology Solutions. “AI and cognitive computing will enable real-time personalisation, using the customer’s engagement patterns and sentiment analysis, impacting buyers’ end-to-end journey.”

What he means is, till now your retailer largely suggested products based on your purchase history—if you bought two bags, it would suggest more of the same. But with AI, it can analyse the bags you bought, and suggest trousers or shoes to match. Amazon, for example, has for some time been using an algorithm to personalise your shopping experience and recommend products such as e-books on Kindle or backpacks on the website.

But where there is opportunity, there is competition. A retailer looking for AI technology has three options: buy a pre-built model sold by large tech companies like Google or Microsoft; approach a developer or third-party company who can customise such a model; or opt for a company like MSD that offers fully personalised AI solutions.

Thus, MSD’s competitors include tech behemoths such as Amazon, Google, IBM, and Microsoft, who have developed core AI capabilities and neural networks that can also be used for retail. For instance, Google provides retailers AI solutions through its Google Cloud Platform.

Besides this, MSD also has to take on smaller companies such as U.S. software development firm for e-commerce Fluid, who use the core technology of larger companies and build a model tailored for retail. Fluid used IBM Watson’s technology to develop conversational AI for retailers. American outdoor clothing brand The North Face is one of Fluid’s customers. IBM has since acquired Fluid.

Smaller companies can offer customised products because large players have ‘open sourced’ the core technology. In other words, they give out ‘keys’, called APIs (application programme interface), of their neural networks. Developers use these ‘keys’ to access the network and develop their own products.

Therein lies the rub for MSD, which offers personalised products for retail clients. Shailesh Ghorpade, managing partner and chief information officer of technology fund Exfinity Venture Partners, says this open-source AI has hurt MSD, but adds retailers “can only go so far with an open-source solution”. He says customisation and understanding the unique challenges of each retailer will be the key.

MSD’s Asokan says customisation is Vue.ai’s main selling point. However, the company found more takers abroad because Indian retailers such as Myntra largely rely on open-source technology to develop in-house products. This was one of the main reasons why MSD moved to San Francisco.

“This is not a typical SaaS [software as a solution] model. We don’t offer a software and tell customers, ‘Go do whatever you want with it’,” Asokan says. “Moving to the U.S. made sense from a business perspective or otherwise. The biggest were not investing in Indian companies. Not in core tech companies. Indian companies have not been known to make core tech breakthrough. So unfortunately for us, that was a stigma. Also, we are working with some big, classic U.S. retail clients, and it made sense to move.”

Retail solutions might be MSD’s identity now, but it wasn’t the plan when Asokan and Chandrasekaran conceptualised the company in 2013. Their ideas were grander. Chandrasekaran is a neuroscientist by training and was part of Stanford University’s ‘Brains in Silicon’ team that built a brain simulator called Neurogrid. Until 2014, Asokan was heading a team at Intel Labs in the U.S. that worked on the future of technology. The confluence of their ideas led them to begin working on AI technology. “I study the brain, its physiology,” he says. “A lot of the AI you see is based on a superficial understanding of how the brain works. The brain has a lot more to it, and the understanding of it will only come if you have the domain knowledge.”

They, along with Costa Colbert—a neuroscientist Chandrasekaran knew from his Stanford days—began developing a neural network that wasn’t simple predictive AI, but a newer generation with the ability to learn from experiences in businesses such as gaming and robotics. So they left their jobs and moved to India. “In India, there weren’t any AI startups at the time; hence, a lot of investors were interested,” Asokan recalls.

But MSD’s investors also had a suggestion: focus on retail. And that led to Vue.ai. The company’s emphasis on core technology and the neural network helped separate it from other AI startups in India such as Bengaluru-based chatbot maker Niki.ai and banking-focussed Hyderabad firm Arya.ai. With Vue.ai being a success—Asokan says she is hoping to be “sitting on a war chest” from revenues from retail by the end of 2017—the trio are looking to expand MSD to other industries. Colbert is now senior vice president at MSD Labs, the company’s research and development wing.

“MSD is not about retail. Vue.ai is about retail. That is why we have created a separate branding,” Asokan says. “I have my eyes on robotics. I’m not saying I’m going to build a robot. But with the kind of software we have, I think we can get into the robotics industry. Health care, definitely, is fair game for us.”