What does it actually take to build AI for 70 crore users?

Vikram sits down with Rahul Chowdhury - co-founder and CTO of PhonePeto talk about how India's most scaled fintech is approaching AI. Not with hype or a top-down mandate, but with a quiet, deliberate, engineering-first philosophy that started four years ago with a small team focused on making developers happier.

Rahul shares the inside story of PhonePe's AI journey from building their own LLM gateway and Agent Hub, to launching AI search with Microsoft, to betting on on-device models for privacy and cost. And it ends with the biggest idea of all: India's DPI stack has spent a decade making data AI-ready.

The opportunity now is to use it to build the bank branch of one — truly personalized financial products for every Indian.

If you're a founder, engineer, or product leader trying to understand where India's AI story is really headed, don't miss this.

What you'll learn

🔹 Why PhonePe avoided output metrics in year one of AI and why it worked

🔹 How to build an AI culture without a top-down mandate

🔹 What an LLM gateway is and why every scaled company needs one

🔹 Why on-device models are the right bet for consumer AI in India

🔹 How DPDP will reshape how companies think about AI and data

🔹 Why India's role in global AI is in applied AI — not foundational models

🔹 How DPI × AI creates the opportunity for hyper-personalized financial products

Chapters

0:00 Intro & who is Rahul Chowdhury
02:30 PhonePe's AI journey: tinkerers to transformers
06:00 The DevX team: why developer happiness came first
10:00 Don't rush into AI — the engineering first mindset
14:30 Building the LLM gateway & data stack
18:00 Agent Hub: PhonePe's internal marketplace of agents
22:00 AI Search with Microsoft & on-device models
26:00 Why India needs edge models, not foundational ones
30:00 DPI × AI: the bank branch of one
34:00 Conclusion