Amazon bets big on machinelearning tech
Online retailer Amazon India is fast increasing the adoption of machinelearning technology in order to cut product returns, improve the speed and accuracy of product deliveries, provide more relevant search results and improve efficiency in other areas of its business.
Amazon is one of the global leaders in artificial intelligence and machine-learning technologies, along with the likes of Google, Facebook Inc. and Microsoft Corp. In India, the company has been relatively slow in using machine learning, mostly because it launched in the country only in June 2013.
Now that it’s accumulating increasing quantities of data on customers, the company is increasing the adoption of machine learning in its business.
“There are a lot of problems that are India-specific where you need machine learning,” said Rajeev Rastogi, director, machine learning at Amazon India (Amazon Seller Services Pvt. Ltd). “One example is address quality. In some of the Western countries, you can catalogue every possible address. But in India, addresses are very unstructured. Understanding addresses better so that when we get an address, we know whether it is complete and whether we can deliver to that address, break it down into neighbourhood, locality, building name. That’s a huge space where machine learning is useful in improving address quality so that we have higher success rates in deliveries.”
The US-based company, which operates in 12 countries, is also exporting some of its machine learning initiatives in India to other markets.
“One of the innovations that we focused on for the India market is considering delivery speed when ranking search results. This innovation is now being considered for other markets as well,” said Rastogi, who is one of the most coveted artificial intelligence experts in India.
Apart from product delivery and location-related uses of machine learning, Amazon is also using the technology in things such as standardising sizes in fashion products and making search results more relevant.
“In softline categories like shoes and apparel, brands have different sizing conventions. A size 6 in Nike may be different from a Reebok’s. One of the problems we’ve had is that a lot of returns happen because people buy the wrong product sizes. We’re using recommendation technology to try and recommend the right product size based on shopping history,” Rastogi said.
BENGALURU: