Keshav Kumar: Making quick commerce personal, predictive, and precise

Keshav Kumar, Head of Engineering, bigbasket.com, first encountered Artificial Intelligence (AI) as he sat in a classroom at Visvesvaraya Technological University learning about neural networks.

“We had a six-month course that equated how neurons in the brain work to how neural networks function. It was fascinating and planted the first seeds of curiosity for me,” Kumar recalls.

Back then, access to infrastructure and data in India was limited. But what Kumar lacked in resources, he more than made up for through brainstorming sessions and problem-solving discussions.

These early days would be the catalyst for Kumar’s professional journey. At NetApp, he began to find practical ways to implement data, applying statistical techniques to uncover insights. AI, Machine Learning (ML) and data had transformed from classroom subjects and theoretical discussions to mission-critical tools powering global scale.

When Kumar joined Prime Video, which was operating in 192 countries, he was tasked with making content truly accessible across languages and cultures. “We had to think deeply about language localization—how a movie made in Korean could resonate with a Telugu or Hindi audience, or how a Tamil film could find an audience in Japan. Subtitles, dubbing, and personalization at this scale weren’t possible without AI and ML,” he says.

It was during the onset of the pandemic and the physical shutdown of studios that Kumar was able to use his experience and expertise to solve problems on a global scale - to keep content flowing for millions of people stuck at home. That was when he knew: AI was not just a support function; it was a force to be reckoned with.

Bringing control to the chaos of quick commerce

With quick commerce emerging as the defining trend of India’s retail landscape, AI is no longer a peripheral tool; it is the invisible engine powering the entire ecosystem. From the way customers search for products to how deliveries are planned, routed, and fulfilled, AI/ML has permeated every layer of the value chain.

“AI gives you a 360-degree value add in all parts of the supply chain,” Kumar explains. On the customer-facing side, it is transforming the experience itself—enabling new ways to search, chat, and shop. Whether it is conversational queries like ‘suggest me a health product for a fitness enthusiast’ or recipe-based searches that generate a full ingredient list for biryani, AI is reimagining how people interact with a platform. It also reduces costs by innovating in customer support, automating responses, and predicting needs before a query is even raised.

Behind the scenes, AI’s impact is even more pronounced. In planning, machine learning models help forecast demand and capacity. In supply chain management, algorithms determine optimal network design, generate dynamic delivery routes, and decide when to assign one executive per order or club multiple orders together.

The final layer is data interpretation. Here, AI not only extracts insights but also translates them into actionable strategies for vendors, sellers, and partners, creating a more responsive ecosystem.

The engineer’s checklist: operation, innovation, foresight

At BigBasket, engineering is the backbone of the business, and leading it is a 320-member team that spans development, security, infrastructure, DevOps, data management, and ML engineering.

Kumar describes his role as Head of Engineering as “threefold”. The first is ensuring seamless operations. “Unless our apps and systems run predictably, the business doesn’t grow. My primary responsibility is to make sure every customer who shops on BigBasket has a frictionless experience,” he says.

The second is driving innovation. From new features that bring in more customers to encouraging higher-frequency shopping and larger basket sizes, engineering decisions directly impact growth. The third is organizational foresight—investing in the right technology bets, scaling architecture, and building systems robust enough to handle the rapid expansion of India’s quick commerce market.

Adding AI to the cart

At BigBasket, the most visible impact of AI is on the shopping experience. Personalization ensures that what matters to one customer may look entirely different for another. “How do you discover products better? How do we make search more intuitive? How do we localize experiences?” Kumar asks. 

The second critical area is demand planning and fulfillment. AI-driven forecasting determines how much stock each store should hold, how frequently it should be replenished, and how many riders are required at different times of the day. This precision ensures timely orders, drives efficiency, and reduces the cost per order.  

The third application lies in anomaly detection across BigBasket’s distributed network. With hundreds of stores operating in multiple cities, disruptions are inevitable—from floods to strike or supplier delays. ML models flag anomalies in real time, enabling teams to react quickly.

Finally, AI plays a role in developer productivity and infrastructure management. From generating boilerplate code to automating code reviews and provisioning infrastructure, AI ensures systems remain resilient as order volumes grow.

At BigBasket, one of the most impactful applications of machine learning lies in demand planning. “You’re talking about 20-30,000 SKUs, as well as individual SKUs across 1000s of stores, each with their own local flavor. Now add in seasonality, even if you just multiply 20,000 with 1000 stores, you’re looking at extremely complex metrics”, says Kumar.

AI has also transformed customer support. Analysis revealed that a large portion of calls were “anxiety calls”—queries about order status or refunds rather than genuine complaints. By deploying GenAI-driven natural language flows, BigBasket has automated nearly 80% of such conversations. 

Trusting the AI process at bigbasket.com

For BigBasket, the first AI-related hurdle was the availability of quality data. The second challenge was more cultural—most companies are used to deterministic software, where algorithms succeed or fail. Machine learning, by contrast, is probabilistic. Building stakeholder trust in such systems is critical, especially when 1–2% of cases may still go wrong and require human override.

The third challenge lies in balancing GenAI’s promise with reality. While demonstrations look impressive, organizations must grapple with constraints such as hallucinations, output quality, and the hard question of whether the investment justifies measurable returns.

Curation, collaboration, and customisation with AWS

BigBasket’s technology backbone rests entirely on AWS, a partnership that has been central to its scale and innovation. The collaboration goes beyond infrastructure support to early access, insights, and hand-holding on emerging technologies. From services like SageMaker to newer tools such as Zero ETL and DMS, BigBasket has often been among the early adopters, with AWS providing guidance at every step.

In the fast-evolving GenAI landscape, this partnership becomes even more valuable. “There’s no week without a new release, but not every improvement is meaningful for us. What AWS does well is curate updates that matter, helping us focus on areas that can create impact,” Kumar explains. This curation enables BigBasket to experiment with confidence, knowing they have AWS’ expertise to lean on—even when outcomes are uncertain.

The collaboration has been particularly useful in areas like personalization, where AWS’ inputs and suggestions have helped BigBasket navigate complexity and deliver customer-centric solutions.

For Kumar, who is also an Amazon alumnus, the alignment is natural.

“AWS has genuinely been a great partner. It's just beyond consumption. They engage us in their roadmap building. They show us previews of what they're about to build. They're even open to us doing architectural deep dives. So I think it's been fantastic.”

Staying aligned with AI

Staying updated in tech, Kumar believes, is about curiosity and discipline. He follows leading engineering blogs, research papers, and forums like Hacker News, Reddit, and Arxiv. Social media, too, plays a role—his feed is largely tech-driven. He also tracks updates from big players like Amazon, Google, Meta, OpenAI, and partners such as AWS or Databricks. 

Outside work, Kumar is passionate about wildlife photography and often drives to sanctuaries to capture moments in nature. A self-confessed food enthusiast, he loves exploring cuisines but admits nothing compares to a good Indian biryani. “I’ve tried everything, but nothing beats Indian,” he says with a smile.