Sumeet Gadodia · Mumbai, India

I follow the problem until it becomes a product.

Product leader and builder. At Aza Fashions, I lead practical AI work across commerce and operations. Outside work, I turn questions into products people can try.

Portrait of Sumeet Gadodia
A product person, in progress
PRODUCT DISCOVERY✳AI IN THE REAL WORLD✳POST-ORDER EXPERIENCES✳BUILDING FROM ZERO✳PRODUCT DISCOVERY
01 / Currently building

Ideas with working parts.

I like the point where a vague idea turns into a screen, a workflow, a test — and eventually something usable.

Consumer AI · India

Nirmit.life

What if making a video felt as natural as writing what you want to say? Exploring a photo, a voice, a script and a way to speak to people in their language.

Creating for India, one experiment at a time

Conceptual visual · Product in development

B2B AI · Support

JoshFast

AI-first customer support for modern teams. A system where answers, playbooks, routing and human judgment are designed together from day one.

Customer conversation✳
Can I change my delivery date?
Check policy → identify exception → ask the right team
KnowledgePlaybookHuman approval
AI that knows when to involve a person

Conceptual visual · Product in development

02 / Notes from the workbench

My thinking,
in the open.

The interesting part is rarely the feature. It's the decision behind it.

A few questions I've been working through while building products, testing models and debugging the less glamorous parts of customer journeys.

01

Which video model survives a real use case?

For Nirmit, I compare identity, lip sync, motion, language, latency and cost with the same input. A good demo clip alone is not enough to choose the default.

02

Could a support team build an AI agent like a coding agent?

In JoshFast, a team describes a situation; Gemini helps turn it into a playbook, expected reply and owner team. Unclear decisions surface in a focused drawer for a human to settle.

03

What does “good enough” mean for catalog AI?

Enrichment needs checks that reflect how people shop. I tested models, trimmed unnecessary work, and designed an AI review followed by human verification where errors matter.

04

When a date is wrong, which system changed it?

A delivery promise can be rewritten by several services. I traced each change, then used an AI audit to point the team to the specific mismatch and its reason.

05

Can you test a market before translating everything forever?

For an Arabic shopping experience, a one-time AI translation and a controlled experiment gave the team a practical way to learn before committing to a costly ongoing approach.

Mumbai street scene at golden hour
Build for the people in the context they're actually in.
03 / Product work

Problems I've spent time on.

Selected work across luxury commerce, service operations and the systems behind a good customer experience.

Aza Fashions / AI in commerce

Taking AI out of the demo and into the daily workflow.

My AI work at Aza spans customer support, catalog and merchandising. The question is how each system makes a useful decision, catches a bad one and knows when a person should take over.

Customer support copilotSales agentsCatalog intelligenceVisual merchandising & taxonomyVoice of customer
From business question to usable system
01 / UnderstandWhat is the customer or team trying to do?
02 / ActAnswer, enrich, classify or route within the right context.
03 / CheckEvaluate confidence, exceptions and business rules.
04 / LearnHuman review where judgment changes the outcome.
AI handles the volume. People control the risk.
Indian luxury fashion browsing scene
Aza Fashions · Discovery

Finding the right thing among thousands.

Search, filters, suggestions, colour relationships and the small decisions that make a large catalogue feel navigable.

Aza Fashions · Post-order

“Where is my order?” is a product problem.

Delivery promises, courier orchestration, cancellations and exceptions — viewed as one journey from the customer's side.

Aza Fashions · Trust & risk

Should every customer see the same checkout?

Exploring COD controls, partial payment and customer behaviour without treating trust as a single yes-or-no rule.

Onsitego · Service operations

What happens after a device breaks?

Claim, pickup, repair and return. I followed field teams, mapped the real workflow and shaped tools around their work.

Aza Fashions · Search quality

One bad result isn't the whole search engine.

Weekly audits and conversations with merchandising helped distinguish genuine exceptions from a broader quality signal.

Across teams · Leadership

Give people the problem, not just the answer.

I've mentored PMs through ambiguous roadmaps, prototypes and experiments. The goal is to help them own the next decision.

04 / How I approach it

A few working rules.

These came from shipping, watching things break, and changing my mind.

01

Start with the awkward exception.

The edge case often reveals what the system really needs to do.

02

Make the idea testable early.

A rough prototype can settle a debate faster than a polished plan.

03

Know when AI should stop.

Build a useful handoff, an evaluation loop and a way to see why a decision was made.

An evening in MumbaiMumbai, where I call home
05 / The path here

What’s the next harder problem worth solving?

That question has taken me across enterprise software, commerce and the systems that keep a customer promise. Fourteen years in, I’m still following it.

Zepo

Where I found product management.

Building ecommerce tools for sellers led me into logistics too, with Zepo Couriers. The operational problem was often bigger than the software screen.

Onsitego

Learning to build behind the scenes.

CRM, support, partner workflows and field operations taught me how much of a good customer experience depends on systems people never see.

Aza Fashions

Rebuilding, then scaling.

From the ecommerce platform and discovery journey to post-order experience and profitable growth, the work kept crossing team boundaries.

Now

Making AI useful in everyday work.

At Aza, I lead work on support copilots, sales agents, catalog intelligence, visual merchandising, taxonomy and voice of customer — bringing AI into operational decisions with room for human judgment.

What excites me is technology that quietly removes complexity, helps people do better work and makes a business move faster.

06 / Let's connect

Tell me the problem you're working through.

I'm open to product leadership conversations, AI product and workflow consulting, and helping PMs think through their resume or interviews.

Let's make the first version tangible.

Share the user problem, the current workflow and what you've tried so far. Find me on LinkedIn ↗