His goal has been to make sure that data science has a healthy presence in each of Flipkart’s different business units and tackle some of the most important data problems where data science can be applied. One such goal is towards building a platform that is inclusive and accessible for the next 200 million users or ‘Bharat consumers’ that Flipkart
is eyeing to tap. Majority of these users come from tier-2 and tier-3 cities, towns and rural regions.
Datar is of the view that technologies such as artificial intelligence
(AI) and machine learning
(ML) would play a critical role to reach these regions. This includes optimising the cost of getting the products to the farthest corners of India. This would, in turn, help the sellers, small merchants and Kiranas which are part of Flipkart’s supply chain, pass on the savings to the end customers. These technologies also help in better understanding of the user's needs constantly helping in areas like product discovery, recommendation and search, especially in the context of Bharat. “We have very good language models and language understanding for English, but how do you create the same magic for Indian languages, understanding the kind of spelling mistakes or the ways in which Bharat users express themselves is another key area where I see AI and ML playing a big role,” says Datar, an alumnus of IIT-Bombay who also has a PhD in Computer Science from Stanford University.
In an effort to tap the next 200 million Bharat consumers, Datar and his team are also leading efforts at Flipkart
for deepening its collaborations with some of the top institutes in the country. The aim is to look at real-life research being done at these institutes in areas such as AI, ML, automation and natural language processing to solve some of the most critical industry problems. This involves leveraging Flipkart’s data and platform knowledge to work on India specific e-commerce challenges and help reach online retail to more consumers and sellers.
The company for example recently formed a collaboration with IIT-Patna which is particularly aimed at developing robust machine translation techniques for translating large volume of user reviews written in English to the Indian vernacular languages. Datar says, IIT-Patna has a very good track record in the area of Neural Machine Translation, a technique to translate one language to another, and predict the likelihood of a sequence of words. “We will put some of these technologies out there so that our users benefit from it. I feel very confident that you will see that happening in the next few months,” says Datar.
Flipkart has also collaborated with institutes such as the Indian Institute of Science (IISc), various IITs and IIMs.
The other areas in which Datar and his team are working include planning the supply chain and giving the right kind of insights to the sellers. For instance, they give insights on the number of units of a product a seller should expect to sell on Flipkart’s marketplace. This way, they can make the right decisions in terms of purchasing inventory. Datar says, as the firm has grown, even an opportunity to improve business including forecasting the demand of products by 5 per cent is huge.
Prior to joining Flipkart, Datar worked with Google as a research scientist for over 12 years. At Google, he led various projects namely keyword suggestion tool for advertisers, Google Adwords broad matching and Google News
personalisation. The other projects included ratings and reviews infrastructure and AdSense targeting for social networks. He and his teams were also credited with working on projects which had a big impact on Google’s bottomline. Before Google, Datar had worked with technology
majors such as IBM and Microsoft
in several research areas.
Datar, however says, the kind of challenges and problem statements that he has seen at Flipkart are bolder and bigger due to the nature of the India market. “What differentiates Flipkart is that we are building for India. You'll be surprised to know that many of the algorithms, language and image science models that were built for the Western world by the (Google's, Amazon’s and the Microsoft’s of the world) just do not work out of the box in the case of India,” says Datar. “India possesses its own challenges (like) diversity of languages, and you need to innovate a lot in the context of India.”