Flipkart data scientist on having the sexiest job of the 21st century

Flipkart’s data scientist Mayur Datar, didn’t actually tell us how he felt having what the Harvard Business Review calls the sexiest job of the 21st century. He only laughed and brushed the notion aside when we asked him about it. Maybe he’s profoundly aware of the complexities and responsibilities that lie on the head that wears the data scientist crown. Maybe tags like those don’t matter, after degrees in technology and computer science from IIT Bombay and Stanford University and a professional journey that began with being research intern at IBM and Microsoft, moving on to 12 years at Google, and now extracting intelligence from Flipkart’s data. Pushing the glamour aside, Datar spells out to Tech2 what it takes to be a data scientist, especially to a giant like Flipkart, which eats tons of data for breakfast. We ask him where an ideal data scientist comes from - the business side or IT? Datar believes that while ideal data scientists would come from the IT side, they need very good domain knowledge that the business functions in. “Data scientists need to work closely with business leaders, because they are the ones to tell you the right questions to ask, the answers to which lie within the data. So skill sets of IT and using knowledge from business to know what kind of insights you need from data is the ideal combination,” he shares. Is a data scientist only as good as the data is? Datar nods in agreement. “Companies should definitely put an effort into correctly gathering data, making sure there are no errors there, no data is missing, and also storing it properly and making it easier to analyse. This is what will unlock most value out of the data. A data scientist’s job is made easier if this is done properly,” says Datar. India is slightly lagging when it comes to maturity of organisations around data sciences. Datar explains, “While the Indian industry has caught on well to the open source revolution and has been using products like Hadoop, when it comes to analysing data. But in terms of cleaning up data, harmonising it, we don’t always do that well. Most Indian companies waste a lot of effort there.” He feels if companies put some thought there early on, in terms of designing the system, architecting systems, it goes a long way to improve the maturity of the data, and makes analysing it simpler. “So it’s not just about using Hadoop, it’s about making the data amenable for say, real-time analytics, and we have a long way to go there,” Datar adds. With all this talk around the magic of data science and CEOs’ perceptions around what a data scientist could do for the company, we ask Datar if those desires get in the way of how he, or data scientists, in general, work. “Companies should think of data science as the technology that gives you the last bit of performance improvement or efficiency that is out there,” he replies. Borrowing an example from his own experience with e-commerce, Datar explains they need to be working at maximum efficiency at all times. They need to constantly monitor and improve supply chain, pricing, kind of products they recommend, just so that it becomes a viable and profitable business, especially in the Indian context. So while it may not necessarily conquer all their problems, Datar says data sciences gives companies that competitive edge. Taking the discussion about e-commerce further, we ask Datar how Flipkart is doing things differently in terms of data science. He points out how Flipkart needs an end- to-end view of data, because their customers interact with them on a variety of systems - be it browsing through products, or placing an order, interacting with their fulfillment services like Ekart, and sometimes returning products. “Traditionally, companies collect data from each of these silos and only look at them in an isolated manner. We are trying to look at the entire user flow or path, and trying to make sense by piecing together the stories to get an idea of what the customer is doing, or how valuable they are as customers. So that would be the key difference that I see - unifying data across different touch points,” reveals Datar. On being asked if Flipkart plans to invest in technologies that see much hype these days - Internet of Things (IoT),  Artificial Intelligence and machine learning, Datar says they have started looking in IoT but the cost factor of it doesn’t make it a very scalable option at this point of time and they are looking to apply it only in selective categories or verticals. But, he says, IoT is definitely being looked at to make things more efficient and to avoid wastage in supply chains. He shares that image analysis and deep learning are very hot these days and the company is looking into those areas to improve user experience. Talking about machine learning, he believes it has a whole suite of algorithms which have been around for 10-15 years. “Our job is not to come up with new algorithms but focus on modelling the data correctly and figuring out how to take the data and model it into the right problem to get the best results,” says Datar, signing off, in true data scientist spirit. Watch a video of Mayur Datar, Data Scientist - Flipkart, speaking to Tech2 here.