Tech is reimagining how e-commerce is done: Flipkart's Jeyandran Venugopal

Flipkart has “invested deeply” in technology to serve millions of its customers, says JEYANDRAN VENUGOPAL, Chief Product and Technology Officer at the e-commerce company. In a video interview with ARYAMAN GUPTA, he spoke about how the firm is using generative artificial intelligence and other technologies. Edited excerpts:

What does Flipkart’s technology landscape currently look like?

We are a platform that serves hundreds of millions of customers on a yearly basis. To cater to that level of scale, with the required reliability and security, we have been investing deeply in building hardened platforms. We've followed the traditional micro services architecture, and we have a fairly large Cloud footprint. We have built a multi-data centre [and] multi-Cloud system, so that we can guarantee disaster recovery, business continuity and seamless scaling, while also being very cost efficient. We have fairly advanced, cutting-edge systems, and we leverage open source quite a bit. We work with several vendor partners also to bring whatever is best in class, when we are putting together systems to serve our customers and sellers.

What are some of the areas Flipkart is betting on from a technology perspective?

There are a few trends that we are strongly betting on. If we look at generative AI (artificial intelligence), which has transformed several industries, we see lots of use cases where we could completely reimagine the customer experience. Regardless, we continue to use traditional AI/ML (machine learning), and predictive AI technologies as well, for several problems spanning our entire operations. That continues to be an investment area. Another big trend is around, what we call, video-fication of commerce. This includes aspects of live video commerce, such as when influencers or content creators sell things in a live stream. And there are several other aspects to video-fication of commerce as well. We also continue to invest in technology like AR/VR, immersive commerce, and blockchain technologies. In the long term, we believe that these will all be part of significant reimagining of how ecommerce is done on the internet.

How much does Flipkart invest in its tech stack?

We don't share specific numbers. But we are a technology- and product-led company. We keep that in mind when we make capital allocation decisions. We invest as much as is needed in our R&D (research and development) budget as a part of our overall P&L (profit and loss) to keep innovation flowing.

When we build products, we measure ROI (return on investment) in several ways. It is not only about the immediate here and now. We measure several things like consumer experience scores across. Through our net promoter score (NPS) surveys, we look at things like (what is) top of mind of customers, concentration awareness, and other such marketing and engagement metrics. That is how we think about technology generally.

How is Flipkart using generative AI?

Generative AI is such a powerful technology that the applications are bound to cut across several aspects of business. On the consumer-facing side, the marquee product is what we call Flippi, which is our conversational assistant bot that helps customers in the entire product discovery and shortlisting process. We have seen a good chunk of our monthly active users experimenting with it, and we continue to improve the capabilities of the bot.

We have also launched products to enrich our catalogue itself. For instance, we are using generative AI to realistically render images of products, with different backgrounds and settings. Another product that is seeing really good consumer traction is what we call a style-made product. Using the feature, a customer can pick an item as an anchor product and refine their search further. We are also using generative AI to translate user-generated reviews from English to Hindi, for example, and then summarizing these reviews.

We developed something called a seller information assistant (SIA), which is basically a seller information assistant bot. It is almost like having a personalised customer account manager for sellers. We have also automated the creation of merchandising cards for sellers, which was a manual process earlier.

There are more than two dozen use cases that we have built using generative AI, but those are some of them.

Does Flipkart see any potential in drone deliveries?

We are not looking at drone deliveries per se, especially when it comes to last-mile deliveries. We have been experimenting with drones for things like mid-mile use cases. But at this point, we do not think drone deliveries are something that we can deploy at scale across the country.

Is there any thought on migrating to a completely in-house large language model (LLM, an AI system)?

It is too soon to comment on that. We have, of course, been experimenting with instruction specific, fine-tuned LLMs that were built on top of open source…But we don't have any ambitions or aspirations to build an in-house LLM from scratch. We don't think that is a reasonable way for us to differentiate capabilities for our customers. Because some of the open-source LLMs have become fairly powerful and capable, we don't need to reinvent the wheel on that.

Does Flipkart operate its own data centres?

We have two private data centres. We work with data centre partners and providers, who build and manage the data centre shell. And we lease out large chunks of space within that and build our own flows from a tech infrastructure perspective, which are operated by our own embedded teams within the facility.