Image: Amit Verma
Flipkart’s top executives are enthusiastic about betting the future of the company on Artificial Intelligence (AI). ‘AI for India’, a campaign that founder Sachin Bansal had spoken about in December, has begun in full earnest, helping the $12-billion (by valuation) ecommerce behemoth focus on harnessing advancements in technology.
“At Flipkart, we’re extremely excited about technology. It has been the biggest driving force behind our success,” CEO Kalyan Krishnamurthy told the company’s developers and engineers while kicking off the fifth edition of Slash N, Flipkart’s annual tech conference, in April. Technology has “not only driven Flipkart’s growth, it has also enabled us to disrupt the market in a big way in the last 10 years”, he said. “We strongly believe that our next wave of growth will come from AI-powered solutions, which will be a key enabler in boosting the entire Indian economy.”
It’s something other companies, in India and worldwide, are betting on too. A recent report by Microsoft, an important investor in Flipkart, projects that, by 2021, about $154 billion will be added to India’s GDP, representing a compound annual growth rate of 1 percent, purely by products and services created or facilitated by the application of AI and machine learning and related digital technologies.
“Access to affordable credit isn’t easy in the Indian context. AI and machine learning can change that.”
—Mayur Datar, Chief Data Scientist, Flipkart
“If you cut through the hype, we believe there’s a lot of value that businesses can unlock through the application of AI and machine learning to their operations and business processes,” chief data scientist Mayur Datar told ‘Flipsters’, as the staff members are called.
Datar added that there is a lot of potential to make operations and business processes much more “intelligent”, thereby making India’s industry more competitive, and unlocking efficiency and profitability. “I want to particularly call out fintech because access to affordable credit isn’t easy in the Indian context, particularly for individuals and small-and-medium businesses,” said Datar. “AI and machine learning can change that.” He pointed out that, even in the public sector, these software programs and the techniques made possible because of them can be applied for better governance, urban planning, and “to make sure the benefits from our taxes actually reach the people for whom they are intended”. Also Read: Why Flipkart thinks it needs to win only one category to take the game away from Amazon India
Amazon pumps funds in India to fight Flipkart
Jeff Bezos' relentless pursuit of India
In the Indian context, the need to apply AI and machine learning is even more fundamental, argued Datar, because “access to technology itself is limited in the country”. “Agriculture, health care, education, urban planning, ecommerce—you can pretty much apply AI and machine learning to anything that’s out there,” he said.
In the outside world, advances in technologies are increasingly getting open sourced and made available to a larger community of tech developers, scientists and engineers. And yet, there is a massive amount of work needed when it comes to India-specific problems because the advances in technology that come out of the West don’t take into account conditions specific to India. For example, the presence of background noise when it comes to speech recognition.
Flipkart’s data scientists and engineers have often had to do a lot of customisation and, in some cases, even start from scratch. India’s diversity in terms of culture, language, literacy and varying economic power of individuals has presented challenges thus far and will continue to do so.
However, one advantage of India’s vast population and growing mobile penetration is that a lot of consumer internet companies are sitting on a huge amount of data. Flipkart, for instance, collects 10 terabytes (1 TB is equal to 1,000 GB) of data on visitors, web page sessions and search queries that it gets in a single day during one of its annual ‘Big Billion Day’ mega sales. “We are sitting on a goldmine here,” Datar said. That said, there are hurdles. Among them is the fact that good data scientists are in short supply in India. Another is the expensive need for massive computing power. The number of companies that have such resources is minuscule, said Datar.
Flipkart has a modest data sciences core team comprising 60-odd people. In comparison, Chinese giant Alibaba and its internet rival Tencent, which is a strategic investor in Flipkart, may have twice as many working in just one area within AI. However, Flipkart is singularly focussed on its AI goals. Its cloud has over a million ‘cores’ (computer processor units) and the company keeps upgrading its computers to make them work faster and more efficiently, claimed Datar. It will also soon add many GPUs—graphics processing units from chip makers such as Nvidia, considered to be useful in AI processing—which would improve Flipkart’s “deep learning” capabilities considerably.
On the talent front, “we have a good team of data scientists internally and we are hiring aggressively”, said Datar. The company has put in place a training programme to upskill its engineers and product managers. This is new to Flipkart. Datar also spoke about how Flipkart is applying AI. For consumers, product discovery and personalised search and recommendation engines are areas where new techniques are being tried out. Even voice search is being offered as an option.
Within Flipkart, AI techniques are being applied to design products by analysing what people say in their reviews. The company also uses AI techniques to “moderate” both images and text and “auto populate” its categories.
Fraud detection is a big area where AI techniques are being applied to recognise patterns that might point to potential swindling by sellers and buyers. And a small team led by Flipkart data scientist T Ravindra Babu has improved the process of locating addresses over the last two years. This is just the beginning in the company’s journey to improving customer experience with better technology. “We would like to use our data and technology to build unique solutions that are made for India, and solve Indian problems,” said Datar.