Three homegrown companies are using AI to change the way we shop and communicate
The field of Artificial Intelligence is not new, it has been around since 1957 proper when Alan Turing theorised his famous Turing test. Through years of debate on ethics and morals of the system, one thing that has been unanimously agreed upon is the benefit of using AI to solve real world problems.

Which is what inspired Tanay Dixit, Product Head at Talespin to use it to solve one of the biggest problems faced by offline retailers, customer retention. After spending 20 minutes with a sales assistant at a top apparel brand, he came away disappointed.

“I am the kind of guy that buys everything online,” He says “So it was a big leap of faith for me to step into a store, I walked in there looking for a grey t-shirt and after spending close to 20 minutes with a store attendant, I walked away empty handed. Why? Because they didn’t have it in my size. The initial seed for what would become Talespin was sown there. I realised that big retail outlets don’t really have a system that shows customers what they want when they want it plus it made the salesperson seem ineffective at his job and the brand lost a sale.”

The AI as a POS System

Talespin is an AI and deep learning startup that helps retailers increase their customer retention rate. “The technology in retail stores has been the same for decades, unlike online which have recommendation engines tailored to your taste, when you visit a retail store, they are always dependent on the person that walks up to you to bag a sale,” Tanay says. “All the money that is spent is to make the store look fancy, so this is an industry that has been mostly untouched by technology. When you analyse it, you realise a store must satisfy the three core questions a customer always has What should I buy? is there something similar to this? and is it available in my size? Using technology if we can answer these questions either offline or online through messenger or the store website, you won’t lose out on potential sales”

Talespin is tailored to answer these questions in two ways. The first of which is known as The Endless Aisle, a POS system that taps into a brand’s inventory to help make the customer’s life a bit simpler.

“With endless aisle, what we offer is a system that ties into a store’s inventory. Even if a store is out of stock on an item, the system can help send the customer to a branch which will have it in stock or give the customer a chance to order it for shipment right there and then. They choose what they want, enter their email and residence address, pick a payment system and that’s it, they receive the item they want and the brand walks away with a customer” says Tanay

Customer interactions using natural language

“It’s an intuitive system that can work using natural commands, say for example red kurta under Rs. 3000, it will recognise that input and intelligently fetch results according to the parameters”

Building such a system wasn’t easy, Talespin worked with a data scientist and a fashion designer to achieve their vision.

“We taught the system using images, once it recognised the images, we assign it tags which are divided into various attributes such as collar, sleeve, even shine which become hooks for the program to learn and recognise,” He says

Getting images to use for the system was easy but teaching it the difference between types of material was where the challenge was.

“Once during tests, men’s kurtas were read by the system as women’s wear” Tanay laughs “That is why we had to get a data scientist and a fashion designer on board to help us make comprehensive a database of materials and styles”

Say hello to bots

The second solution that Talespin provides is intelligent conversational bots.

“Thes bots can be plugged into any API the store website has or their Facebook page” says Tanay “The beauty about these bots is that we give stores complete access to them, which means at any time anyone from the store can take over the chat manually, if they feel it’s not communicating properly, ultimately we would love for them to take over completely but that’s not a request that goes down easy in India, we have had brands that have told us that a human assistant can be assigned to multiple tasks, like for instance getting coffee,” laughs Tanay “A bot can’t get you a cup of coffee”

AI as a fashion discovery tool

Like Tanay, Atul Rai, CEO and Co-founder at Staqu set out to solve a similar problem but on a different platform. The app Fashin helps users identify what a celebrity is wearing using just a YouTube video.

“Fashin is a completely automated app powered by our VGrep API (our proprietary AI API suite which we license to various fashion e-commerce companies). It has more than 7 million products enlisted in it along with all the features like image search, video based recommendation, meta tag based filtering, out of stock handling, trend based recommendation and various other cutting edge features integrated in its core,” says Atul “We utilised the best AI technology, to make things as simple and as intuitive as possible. In our knowledge, Fashin is the app which has all the features collectively which various apps in the fashion domain have at the individual level which we believe is unique in itself and solves the core problem of fashion discovery through different means”.

Staqu was set up in 2015 by a team of four people who handle various aspects of the technology within the company, Chetan and Anurag are the software engineers, while Atul brings with him experience on working with AI extensively and Pankaj handles server architecture and development.

Fashin uses a similar approach to Talespin when it comes to teaching the AI differences between several types of clothing.

“If you look fashion from an AI perspective you will find that 70-80 % of the content (data) on any fashion e-commerce platform are in the form of images and after working in the image related AI field for the last 8 years, I have experienced that images are the most powerful but least used data set at commercial level. Even fashion companies use the images as more like decoration stuff. When we entered the fashion AI domain, we targeted these images to extract the meaningful information about apparel. We started accumulating these images to train our AI engine which can learn any pattern, colour, style etc. and further be used in automating various tasks. Currently, we are using our proprietary bi-directional deep learning algorithms to learn various meaningful information from different fashion categories” says Atul “Fashion is one domain which is plagued by discovery problems for users. Consumers spend a lot of time finding out the correct apparel and we believed that we can solve this problem using our varied domain level expertise”

The app also uses AI to float intelligent suggestions using photos taken by the user and acts as a one-stop shop for all the latest deals from the big brands.

“At Staqu Wherever we see images, we see opportunities. We are an AI company and as an AI company, our approach is to target data first and then domain. When you start looking at AI with “domain” perspective” says Atul “it may seem over ambitious but if you start looking at it with “data and technology” perspective you will find that the technology which is able to learn the pattern from fashion image data can also be used to learn the pattern of cancerous cell from medical images data”.

Using AI to communicate

Another application of AI is what companies like Reverie are working on, using machine learning to translate languages to make devices like smartphones accessible among a vast amount of users.

“Our mission is to provide language equality on the internet,” says Vijaynanda Parbhu, VP Client Engagements at Reverie Language Technologies “Through our technology, we make it possible for people to translate, display and input various languages quickly and efficiently. We support multiple languages across multiple platforms” He says

It was this thought process that led Reverie to develop the Indic Keyboard Swalekh Flip, a keyboard replacement for Android and iOS which allows users the freedom to type in 22 Indian languages in addition to English. Reverie is helping the mobile brands in India to comply with the recent BIS mandate which makes it compulsory for phones manufactured in the country to display 22 Indian languages and have input support for English, Hindi and one regional language by July 2017.

“Displaying languages is only part of the problem,” says Bhupen Chauhan from Reverie “The real challenge is displaying it in proper context as per the language, Indian languages possess characters other than vowels and consonants which are used for composing words, which is why typing in a language other than English needs to be taught. It is important that the rendering engine should understand and allow the grammar rules of the concerned language to deliver the appropriate output”.

Reverie’s latest project is for the Indian Government, it has been tasked with localising the Bharat Interface For Money or BHIM app into various regional languages. “The BHIM app was built keeping in mind the needs of India’s local language customers from the very beginning rather than as an afterthought. Since its inception, language localisation is integral to the usability of the app and its language infrastructure has been designed around the complexities of Indian languages,” Says Vijaynanda “We have currently rolled out support for 11 languages plus English with more coming soon”

One of the key areas that Reverie is focusing on is to support manufacturers in making cost effective multi-lingual phones that are available in Tier II and Tier III cities.

“Our technology works across multiple categories of phones,” says Bhupen “We support everything from smartphones to basic entry level feature phones”

The second area that Reverie focuses on is mobile app localisation. With the current jump in mobile usability in India, it has become imperative to cater to local language users as well.

“We feel that device-based apps and mobile web browsers, such as search, system information or the user interface will need to localised for the mass market across the country, our technology is built on a rule based compositional engine which means we have greater accuracy, we can accurately predict typing behaviour and display fonts which are accurate to the script being used” says Vijaynanda