SR. PRODUCT MANAGER — B2B × B2C × ONLINE × OFFLINE

I make sense
of messy
products.

Product × Strategy × AI × Outcomes.
Currently turning complex problems into simple product experiences — from enterprise boardrooms to food-court kiosks.

Resume Projects Contact
SCROLL
to find the solution ↓
Mess→Questions→Insight→Product→Outcome✺B2B✺B2C✺Online✺Offline✺ Mess→Questions→Insight→Product→Outcome✺B2B✺B2C✺Online✺Offline✺
01How I Think

My product brain.

I don't start with features. I start with why the problem exists — then work the mess until it becomes something a team can actually ship.

01

Find the real problem

The stated problem is rarely the actual one. I dig until the "why" stops moving.

02

Connect the dots

Business goal → user behaviour → product → technology → outcome. In that order, every time.

03

Turn ambiguity into action

Messy requirements don't scare me. They're usually where the interesting work begins.

04

Build with AI

I look for where AI changes the workflow itself — not where to bolt on a chatbot.

05

Think in outcomes

Shipping isn't the finish line. Impact is. One metric, one named owner.

…and that's how a mess becomes a roadmap ↓
02My Range

Boardrooms to checkouts.
Factory floors to food courts.

Most PMs pick a lane. I've shipped in all four — which means I've seen how a decision made in a dashboard lands on a shop floor.

B2B

Enterprise

PLM & planning systems for a 35+ store apparel retailer. Compliance SaaS across 3 product pods. Restaurant platforms with 800+ outlets.

B2C

Consumer

Voice kiosks in 500+ locations. A civic app built with a state government. QR ordering that cut order time 40%.

Online

Digital

Luxury-brand checkout flows with an agentic layer. Embedded analytics, rule engines, universal search — decision data where there was none.

Offline

Physical

Production execution on real factory floors — SMV drift cut to under 3%, returning ~600 supervisor hours a year.

03Projects

Projects —
built, challenged & changed.

Messes I walked into, questions I asked, and what came out the other side.

7 client & company projects5 side builds & concepts
A

Client & company projects

Shipped at work — click any card to open the full story.

01

One agent platform. Three very different industries.

B2BAI AgentsGemini EnterprisedevX AI Labs · 2026

The Mess

Retail, apparel-manufacturing and media teams all asked for "an AI agent" — and each meant something completely different by it.

My Question

What does this team actually need the agent to do — and how will we know it works?

What I Did

Lead the design and development of Gemini Enterprise (GE) agents: use-case discovery → agent design → build → evaluation → rollout to client teams.

The Outcome

GE agents taken from discovery to rollout across retail, apparel manufacturing & media.

02

The factory floor was leaking 600 hours a year.

B2BManufacturingOfflinedevX AI Labs · 2026

The Mess

An apparel manufacturer's SMV drift sat at 15–20% — planned production time versus reality — and nobody could see where it went. Supervisors firefought instead of supervising.

My Question

What if the floor could see drift the day it happens, not in next month's report?

What I Did

Floor-level discovery → mapped where time actually leaked → designed production-execution features around the client's existing ERP and POS, not against them.

The Outcome

SMV drift 15–20% → under 3%. An estimated 520–600 supervisor hours returned every year.

03

35 stores. One design-code master.

B2BRetail · PLMdevX AI Labs · 2026

The Mess

A premium apparel retailer running made-to-measure, inventory, demand planning and allocation across 35+ stores — on processes held together manually by 10+ teams.

My Question

What's the one source of truth everything else should hang off?

What I Did

Designed Ops Core — the planning and product-lifecycle layer around the existing ERP/POS estate. One design-code master across made-to-measure, inventory, demand and option-health planning, pre-production and allocation.

The Outcome

Rolled out across 35+ stores in and outside India, automating manual process for 10+ teams.

04

Teaching a kiosk to take your order.

B2CVoice AIRestaurant TechCyntra Labs · 2023–24

The Mess

Lunch-hour queues, order errors, and a restaurant chain that needed throughput without hiring its way out of the problem.

My Question

What's the fastest a hungry human can get from "I want that" to "paid"?

What I Did

Took a voice-enabled self-ordering kiosk from concept to rollout. Then went after food courts with QR ordering — several outlets, one cart.

The Outcome

Live in 500+ locations across India. QR ordering cut order time by 40%.

05

The compliance product that trained its own users.

B2B SaaSComplianceSprinto · 2024–26

The Mess

Three product pods shipping inconsistent experiences, compliance frameworks evolving under everyone's feet, and enterprise customers falling behind on training.

My Question

Why does the product get smarter every release while its users don't?

What I Did

Standardised UX across pods, shipped a rule engine + universal search + embedded analytics, applied AI to the highest-volume audit workflows, and rebuilt training and enablement around the frameworks customers were actually audited on.

The Outcome

Training completion up 45%. Design inconsistencies down 40%. Audit-readiness up across enterprise accounts.

06

800 restaurants. One operations brain.

B2BRestaurant TechPosist · 2021–23

The Mess

International restaurant chains running front-of-house, billing, analytics, CRM and back-of-house on one platform — with delivery work stalling somewhere nobody could name.

My Question

Where does work actually stop moving — not where do people say it does?

What I Did

Used cohort analysis to find the real bottlenecks, reshaped the process around them, and owned a mobile inventory app end to end — UI/UX through launch.

The Outcome

Delivery bottlenecks cut 83%. Inventory app reaching 1,600+ users across 800+ outlets.

07

Complaints that route themselves.

B2CCivicCo-founderKaryavahi · 2016–18

The Mess

Civic complaints in Rajasthan disappearing into paperwork — citizens couldn't report safely, authorities couldn't prioritise what they couldn't see.

My Question

What if a complaint knew where it needed to go?

What I Did

Co-founded Karyavahi with the Dept. of IT, Govt. of Rajasthan. Led a team of 4 building anonymous submission with sentiment analysis, auto-routed to the responsible authority.

The Outcome

Top 3 of 7,000 applications in the state. Pitched to entrepreneurship cells of Stanford, UC Berkeley & Northeastern.

B

Side builds & concepts

Things I made on my own time — to learn, to test an idea, or because a real user needed it.

Side build

Think Like a PM

A gamified product-thinking game for engineers — branching scenarios, an animated PM character and a live leaderboard. Launched on Product Hunt.

Solo build · 2026
AI demo

Willow — multi-agent retail concierge

A Gemini Enterprise chatbot for a concept sneaker store — Dialogflow CX subagents, Vertex AI Search data stores and full architecture docs.

Presented to a CTO · 2026
Side build

Rise & Shine Gym Manager

Gym management web app for a Jaipur training centre — role-based views, slot logic, fee tracking, physio management, built for low-tech staff.

Web app · 2026
Concept

ReBin — doorstep recycling

Problem framing, user journeys and end-to-end design for a pickup app that helps households refurbish, recycle or resell.

B2C design project
Prototypes

Retail prototypes & POVs

Rapid prototypes and point-of-view demos for enterprise retail and D2C clients — PLM, ops platforms, retail-intelligence dashboards.

GitHub ↗
04Skills

Skills —
the kit behind the work.

Methods, tools and domains I use to turn the mess into a product.

Product
Discovery & user researchPRDs & user storiesRoadmapping & prioritisationA/B testingRelease managementGTM & product marketingCompetitive analysisStakeholder management
AI Tools
ClaudeGemini EnterpriseEmergentReplitElevenLabs
Data
SQLMixpanelPower BICohort & funnel analysisMetric designCustomer data analysis
Design
FigmaMiroExcalidrawWireframingDesign systemsUser journeysPrototyping
Domains
Compliance & GRC SaaSRetail & merchandisingRestaurant techApparel manufacturingProcurement
05Try Me

Don't read how I think.
Experience it.

Scenario · You decide, I explain

You're the PM. Your company has three fires:

  • Revenue is falling.
  • Customers are complaining.
  • Engineering wants to build something new.

What do you investigate first?

interesting. here's what I'd actually do →

06Unpopular Opinions

Things I'll say out loud.

01

A roadmap is not a strategy.

+

A roadmap is a sequence of bets. Strategy is knowing which bets you're refusing to make, and why. If your roadmap survives every new piece of information unchanged, it isn't a strategy — it's a queue.

02

More features rarely fix a broken product.

+

Feature requests are symptoms. When a product is broken, adding features is how you bury the problem under new surface area. I'd rather delete three things than ship a fourth that compensates for them.

03

Engineers should be in the room before the solution exists.

+

I was an engineer. The best technical ideas come from people who saw the raw problem — not a finished PRD. Handing engineers a solution to build wastes the smartest people in the building. I even built a game about this.

04

AI isn't a feature.

+

Adding a chatbot because everyone else did is decoration. The real question is whether AI can collapse an entire workflow — remove the step, not decorate it. If it can't, don't ship it.

05

A beautiful prototype can hide a terrible decision.

+

Polish is persuasive — that's what makes it dangerous. I stress-test the decision underneath before falling for the Figma. Shipping isn't the outcome; the metric moving is.

07Currently Exploring

Things I'm obsessed with.

keep scrolling — they move →swipe to explore →
→ live

Product thinking for engineers

Built "Think Like a PM" — a gamified branching-scenario game. Launched on Product Hunt.

→ live

Multi-agent AI retail

"Willow" — a Gemini Enterprise concierge with Dialogflow CX subagents and Vertex AI Search.

→ now

Agentic commerce

Adding agentic layers to luxury retail checkout — where the store talks back.

→ now

Outcome management

One metric, one named owner per process. Making accountability a product feature.

→ next

AI-native workflows

Not AI added to workflows — workflows redesigned because AI exists.

→ always

Building from zero

Zero-to-one products: gym ops software, civic tech, whatever has a real user attached.

→ you?

Your messy problem

Got one of these brewing? Let's break it down together →
08Who I Am

Hi, I'm Kavya.

certified mess-untangler ↓ Kavya Chaturvedi

I'm a Senior Product & Outcome Manager at devX AI Labs who enjoys sitting in the uncomfortable space between:

business problems × user problems × technology

I like problems that don't come with a clean PRD. The ambiguous ones. The ones where everyone has a different opinion about what should be built — because that's usually where product thinking matters most.

I started as a software engineer, co-founded a civic-tech startup with a state government, and have since shipped products for factories, restaurants, retailers and compliance teams. Different rooms, same job: find the real problem, make it simple, make it matter.

RetailCompliance SaaSRestaurant TechManufacturingCivic TechAI
0yrsproduct & engineering
0+industries shipped in
0+outlets on platforms I've owned
0+kiosk locations, concept → rollout
0hrsreturned to one factory, every year

Want the boring version?

DOWNLOAD RÉSUMÉ ↓
09Your Turn

Have a product problem?
Give me the messiest version of it.

No polishing needed · mess welcome
here's how I'd start untangling it →
Clarify the problem→ Form hypotheses→ Data I'd want→ Map opportunities→ Prioritise→ Run the experiment

That's the shape of it — the interesting part is what your specific mess reveals at each step. Send it to me and I'll break it down properly →