Product work, systems, and writing

I make AI products survive contact with real customers.

Led $1B+ device-financing platform in 18 live markets, brought LLM search to Galaxy phones, drove experimentation culture and built the evaluation loops that keep AI features honest after launch. My experience spans across consumer, fintech and B2B SaaS domains.

Read the blogStart a conversation
Aadhar Agarwal on a waterfront promenade at sunset
Building products that can take contact with reality.

Where I’ve worked

SamsungSamsung Research · 2024–26
AirtelAirtel India · 2022–23
BrevoPM → Lead PM · 2019–21

Selected work

Samsung Finance+ · 2024–26 · Restricted case study

Replacing a single lender

Samsung Finance+ relied on one lender. That limited approvals, conversion, and the markets we could enter.

+15%
sales conversion
<30 sec
offer time, from days
10% → 22%
repeat-customer share, in 4 months
Explore case study
All contacts90% matchbot vs agent20%cohortShip

Airtel · 2022–23 · Restricted case study

$1.5M out of the support bill, with CSAT going up

Airtel served 300M+ subscribers on $18B of revenue through costly human support. The brief was simple: reduce cost without lowering the experience customers came for.

$1.5M
saved annually
+10%
tribe-wide CSAT
+15%
agent response speed
Explore case study
All work

Latest writing

· 8 min read

The Bottleneck Was Never the Model — And It Just Moved

The companies pulling ahead are not waiting on a smarter model. They noticed the bottleneck was never the model — and that since they fixed it, it has moved twice more. — Aadhar

agent-harnesshuman-agent-collab

· 5 min read

Agentic Memory Taxonomy

How we structure context across short-term reasoning and long-term storage in agentic loops. Defining the boundaries between RAG, vector stores, and persistent world models. — Aadhar

agentic-workflowsmemory-design

· 7 min read

RLHF for Product Managers

Reinforcement Learning from Human Feedback isn't just a technical training phase; it's the new "User Research". How to structure human feedback loops to align agentic outputs with product outcomes. — Aadhar

rlhfpm-craft

The Builder’s Log

In the lab

In progress

NextBite

A meal generator that explores how small models handle hyperlocal preference.

Try NextBite

Get in touch

For a role, a case walkthrough, or a useful comparison of notes.

Email Aadhar