Why Zepto's Aadit Palicha Turned Down Stanford to Deliver Groceries
Aadit Palicha, co-founder of Indian quick-commerce grocery startup Zepto, discusses how he turned down Stanford at 17 to build a 10-minute grocery delivery company during COVID-19, starting from a WhatsApp group in Mumbai. He explains how relentless customer focus, a pivot to dark stores, and obsessive supply chain optimization turned a neighborhood delivery experiment into a billion-dollar platform employing over 200,000 people.
Summary
Aadit Palicha, co-founder of Zepto, recounts how he and co-founder Kaivalya (KB) began delivering groceries to neighbors in Mumbai via a WhatsApp group during the first COVID-19 lockdown in 2020, when they were 17 years old and had deferred their plans to attend college in California. Rather than immediately dropping out of school, they spent roughly a year validating the business, only committing fully once they reached around 10,000 orders per day and a 60-70 crore GMV run rate, at which point they also had investor interest and a term sheet in hand.
The original product, called Kiranakart, involved doorstep delivery sourced from local mom-and-pop stores. Through direct conversations with customers at their doors, Aadit and KB discovered that no existing player — including themselves — was truly satisfying customers on the four key dimensions of speed, quality, selection, and price. This led to a pivotal decision to take control of the customer experience by building their own mini dark stores, starting literally in KB's apartment. The neighborhood served by this first dark store generated three to four times the volume of all other neighborhoods combined, validating the model decisively.
Aadit explains that the 10-minute delivery concept emerged from a first-principles thought exercise inspired by Brian Chesky's 'five-star to seven-star' framework: strip away all constraints, imagine the most extreme positive customer experience possible, and then work backwards to figure out how to make it feasible. Critics at the time argued that owning dark stores was not financially viable, but Zepto's approach generated far higher throughput than anyone anticipated, which in turn dramatically lowered per-unit costs — an outcome that was impossible to forecast without first achieving that level of customer delight.
On the operational side, Aadit emphasizes that Zepto is fundamentally a logistics and supply chain company — an 'atta-dal-travel company' in his words — not a software app. The company now employs over 200,000 people including pickers, packers, truck drivers, and delivery partners, and has built industrial-grade automation in its backend supply chain. Zepto is today the largest seller of fruits and vegetables in India on its platform, sourcing millions of units per week directly from farmers across the country.
Zepto has also developed a substantial advertising business worth hundreds of millions of dollars in ARR, where major brands like Coca-Cola, Pepsi, and Nestlé bid on search keywords. AI and machine learning are deeply embedded in operations: ML algorithms now run supply chain forecasting for millions of units per day without human intervention, and generative AI tools power the ads platform and internal engineering productivity. Aadit notes that Zepto has cut nearly all third-party SaaS and managed services spending by building in-house.
Looking ahead, Aadit envisions Zepto becoming India's homegrown equivalent of Amazon — a large-scale urban grocery infrastructure platform operating across the top 40-50 cities, enabling a new generation of consumer brands to be built on top of its platform, much as Walmart and Costco enabled brand creation in the US decades ago. He attributes his own rapid personal growth to surrounding himself with experienced, senior leaders who joined early and from whom he shamelessly asks even the most basic questions.
Key Insights
- Aadit argues that Zepto did not pivot to the dark store model by benchmarking competitors, but purely by talking to customers at their doorsteps — customers consistently said no existing player, including Zepto itself, was meeting their needs on speed, quality, selection, and price.
- Aadit claims the 10-minute delivery concept came from a Brian Chesky-inspired first-principles exercise: remove all constraints including laws of physics, define the most extreme positive customer experience imaginable, and only then work backwards to figure out how to make it operationally possible.
- Aadit states that nailing customer experience produced a supply chain cost advantage that critics never anticipated — because dark stores generated far higher throughput than anyone forecast, all per-unit costs fell dramatically, making the model financially superior rather than inferior.
- Aadit did not drop out of college immediately; he and KB waited until they had roughly 10,000 orders per day, a 60-70 crore GMV run rate, and an investor term sheet before fully committing — arguing founders should have proof of concept and early PMF before taking that leap.
- Aadit says Zepto has cut nearly all third-party SaaS and managed services spending to zero by rebuilding tooling in-house using AI, and that its engineering team of ~500 is now able to build significantly more with the same or smaller headcount than before the AI boom.
Topics
Transcript
[0:00] If you remove all constraints and you just remove all the laws of physics and you just think from first principles, what's the most extreme positive customer experience you can give and you start from there and then you work backwards from how can I make that possible. So audit everyone in this room knows Zepto almost everyone uses the product many of them probably every day. I don't know how many people know the true story of how Zeppto got started. So I thought we'd start there. How did you end up [0:31] starting this incredible company? >> Well, firstly, you know, Jared, thanks for having me and it's a privilege to be here at YC Startup School.…
Full transcript available for MurmurCast members
Sign Up to AccessMore from Y Combinator
YC Debates: Should We Slow Down AI?
Jacob Coxin argues governments should substantially intervene to slow frontier AI development, citing rapid capability growth (superhuman AI researchers by 2027), alignment risks, and misaligned corporate incentives. Robin Hanson counters that typical technological adoption patterns suggest slower real-world impact, that regulation is dangerous without concrete problems, and that liability-based approaches are preferable to centralized global regulation.
How To Build The Future: Eric Glyman, Co-CEO of Ramp
Eric Glyman, co-CEO of Ramp, discusses how the company evolved from Parabus to build a $60 billion financial operations platform, and how Ramp is leveraging AI agents and internal tools like Inspect, Research, and Glass to transform organizational productivity and reshape the future of work.
Grok Bot: What Changes When You Can Build AI Teammates
Lauren Tan and Roman Ugarte discuss Grokbot, a platform for creating AI teammate agents with their own identities and tool access that can work autonomously. The conversation covers how AI agents are transforming software engineering and non-technical work, with Lauren showcasing her 4,000+ PRs/month workflow using specialized bots in Slack and describing the shift from execution-constrained to execution-abundant organizations.
Grok Bot: What Changes When You Can Build AI Teammates
Lauren Tan and Roman Ugarte discuss Grockbot, a product enabling users to create AI teammates with their own identities and capabilities. They share how AI agents are transforming software engineering and business operations, demonstrating workflows where engineers ship thousands of PRs monthly by leveraging specialized bots integrated into team communication tools.
Is AI Now Good Enough To Cure Cancer? | YC Paper Club
The YC AI for Bio Paper Club event explores how artificial intelligence is transforming cancer research and biological discovery through foundation models, personalized genomic analysis, and autonomous research systems. Speakers demonstrate that cancer is becoming a data problem solvable through AI agents, DNA language models like EVO2 are enabling efficient protein and genome design through few-shot prompting, and validated lab results show these approaches are moving from theory into clinical practice.