1
Sketch first
I draw the flow, the dead ends and the edge cases before I write a spec. Engineers get fewer surprises and designers get a clearer brief.
Open to APM and PM roles · Bengaluru
Moulshree Bhutra
Product Manager.
I sketch the flow before I write the PRD. Founding APM at Vudio.ai, where I helped a video tool grow into a marketing suite for 1,000+ D2C brands, including Shoppers Stop and Godrej.
0 to 1 productsAI productsOnboarding and activationDesign systems
Moulshree, wrapped
Vudio.ai, April 2024 to presentProducts4
0 to 1 launches I contributed toViews10M+
widget views driven for clientsBrands1,000+
D2C brands on the platformARR7x
$30K to $200K in two yearsTeam3
designers hired and ledTurnaround69%
faster, from 13 days to 4
About
From pixels to product.
I studied computer science, then fell for product design and taught myself the craft. I started in design, with internships at Find Firms and Crux Intelligence, freelance work on Contra, and a role at LeadSquared where I designed the My Trips module and a mobile design system for a field sales app.
In April 2024 I joined Vudio.ai as a founding APM, working directly with the founder across product, customer success and design. We took Vudio from a video commerce tool to a full-stack marketing suite for D2C brands. I owned PLG and funnel optimisation, contributed to four zero-to-one launches (AI Studio, Content Library, Social Media Posting and DM Automation), and hired and led a pod of three designers.
I like the unglamorous parts of a product: onboarding, activation, the dashboard a customer opens late at night to decide whether to renew. I build things so the proof is visible, in numbers a merchant can see.
Now I am looking for a product team where I can keep owning problems end to end, alongside people who push my thinking.
- Based in
- Bengaluru, India
- Experience
- 3+ years in product and design
- Education
B.Tech, Computer Science, Poornima College of Engineering, Jaipur, 2017 to 2021- Fellowship
Product Management Fellowship, NextLeap (2024). Top 3 of 200+ fellows.- Domains
- B2B SaaS, D2C and e-commerce, AI products
- Looking for
- APM or PM roles, in-office or remote
2
Own the whole loop
Discovery, specs, design, launch and the numbers afterward. At a small startup there was nobody to hand it to, so I learned all of it.
3
Show the proof
A dashboard that lets a customer see their own return beats a sales deck. I build the measurement into the product.
Product
DiscoveryPRDs and user storiesRoadmappingActivation and onboardingA/B testingGo-to-marketJiraGoogle Analytics
Design
FigmaDesign systemsPrototypingUX research
AI and building
Prompt frameworksGenAI product flowsClaudeGeminiChatGPT
Where I have worked.
Two companies, one thread: make complicated software feel simple for the person using it.
Associate Product Manager
Vudio.ai
Helped Vudio evolve from a video commerce tool into a full-stack marketing suite for D2C brands, scaling to 1,000+ clients including Shoppers Stop, Godrej and Littlebox India.
1Learn the customerWorked across product, customer success and design, so I heard merchants' problems first hand.
2Own the productJoined as a founding APM. Owned discovery, specs, design flows and launches end to end.
3Build the teamHired and mentored three designers and built a new design system.
What I owned
Scope of the role
Funnel optimisation for Shoppable Widgets
PLG: self-serve activation, Shopify App Store launch, usage-based upgrade triggers
Platform expansion: AI Studio, Content Library, Social Media Posting, DM Automation
Internal AI workflows: customer feedback analysis, automated client emails
Impact
Things I shipped or movedTurnaround13 → 4 days
Self-serve onboarding
Built self-serve onboarding and redesigned the whole Shoppable Widgets module, the company's core revenue product. Activation improved by about 74 percent and turnaround time fell about 69 percent.
ARR7x
Revenue in two years
Contributed to ARR growing from $30K to $200K through four zero-to-one launches: AI Studio, Content Library, Social Media Posting and DM Automation.
0 to 1 launch+5% ARPU
AI Studio
A GenAI video creation tool for D2C brands. I led it from zero to one and owned the prompt framework, API integrations and UX. ARPU rose 5 percent among fashion brands.
Conversion~10%
Analytics dashboard
Gave clients visibility into widget performance across engagement and sales data, and improved free-to-paid conversion by about 10 percent.
Clients1,000+
Scale
Built multiple products and grew them to 1,000+ clients, including large Indian brands such as Shoppers Stop, Godrej and Littlebox India.
Team3
Design pod
Led Vudio's design pod, hiring and mentoring three designers and building a new design system for better usability and visual consistency.
Reach10M+
Widget views
Optimised widget and video placement across high-intent, high-traffic pages using Shopify analytics.
Product Designer
LeadSquared
Designed the My Trips module and a mobile design system for a B2B SaaS field sales app.
What I designed
Two pieces of the app, plus the team around themYennic DS
A mobile design system rolled out across all app modules, for a consistent look and fewer points of friction. It helped users complete tasks more easily.
My Trips
A complex field sales module, designed end to end: user research, wireframes, flows, information architecture and prototype. It lifted daily active users by about 10 percent.
Working with the team
Worked with product managers, engineers and senior designers to streamline the design process and speed up iteration.
Also:
Freelance product and design, Contra (2022 to 2024)
Product design and research intern, Crux Intelligence (2021)
Product design intern, Find Firms (2019)
Freelance product and design, Contra (2022 to 2024)
Product design and research intern, Crux Intelligence (2021)
Product design intern, Find Firms (2019)
Recognition
Outside of workTop 3 of 200+ fellows, NextLeap PM Fellowship, 2024
Winner, Popquest International Design Challenge 2021
NID Ahmedabad prelims, AIR 10, 2021
Top 2 global exchange student, Denmark, 2016
Rationale
Customer feedback in. A ranked roadmap out.
An AI copilot for product managers. It groups messy customer feedback by the need underneath, ranks themes by the revenue at stake, and drafts the PRD. Every judgement stays visible and editable.
Concept projectAI-nativeLive demoInvented data
Saved example output
Gut check
Think like the PM. Rank the themes by what you would build first, then see how the arithmetic ranks them.
Feedback inbox
Themes, ranked
score = ARR at stake × confidence ÷ effortPRD draft
Show your work on every roadmap decision.
Feedback arrives from calls, tickets and renewals. The product manager has to turn it into a ranking that sales, engineering and the CEO will accept. Rationale does the tedious grouping and drafting with AI, and leaves each judgement call visible and editable.
Three things go wrong today.
It is scattered
Feedback sits in call notes, support tickets and CRM fields. Nobody sees all of it at once.
The loudest account wins
The most recent or most senior customer shapes the roadmap, whatever the pattern across accounts.
It is hard to defend
A ranking in a spreadsheet has no evidence trail, so every priority call turns into an argument.
These are hypotheses from how B2B teams commonly work. The first thing I would do is interview eight product managers and check them.
Two people share one decision.
Primary · hypothetical persona
Priya, product manager
Owns the roadmap at a 200-person SaaS company.
- Needs: group feedback fast, explain the ranking.
- Fear: missing a pattern, or ranking by gut feel.
Secondary · hypothetical persona
Arjun, head of sales
Brings customer requests to product every week.
- Needs: see why a request ranked where it did.
- Fear: a deal-blocking request sitting at the bottom.
Four steps, one human in charge.
- CollectFeedback with customer, plan and ARR attached to each item.
- GroupAI groups items by underlying need and proposes effort and confidence.
- RankPlain arithmetic turns those inputs into a score the PM can overrule.
- DraftAI writes a PRD from the evidence in the chosen theme.
Arithmetic anyone can check.
The AI proposes the inputs. The formula is fixed and shown on screen, so the ranking can be questioned line by line.
$366k ARR at stake×0.9 confidence÷4 effort=82 · rank 1
$15k ARR at stake×0.5 confidence÷5 effort=1.5 · rank 6
Clause search has two requests and ranks last. Enterprise security has three requests worth $366k, all tied to renewals, and ranks first. The reasoning is on screen for sales to read.
What I chose, and what I turned down.
The score is arithmetic
If a head of sales asks why a request ranked low, the answer has to fit on one line.
Turned down: a single AI priority score. It would be impossible to explain.
Every theme shows its evidence
Each theme lists the exact feedback items inside it and colours them in the inbox. A grouping you cannot audit will not be trusted.
Turned down: sentiment scores. They look precise and tell the PM very little.
Revenue gates are flagged
When an item says a renewal depends on a feature, the theme gets a visible flag so it cannot be buried by volume.
Turned down: treating all ARR as equal, which hides a deal that is about to churn.
The page works without AI
It opens on a saved example so a visitor always sees a complete result. Live grouping is optional.
Turned down: an empty screen until the model answers.
One number, three inputs, two guardrails.
North star
Hours from raw feedback to a PRD the team reviews
Today this takes a day or more by hand.
Inputs
- Grouping accuracy against hand-labelled items
- Share of themes a PM accepts without regrouping
- Share of PRD lines kept after review
Guardrails
- Invented customers or numbers in a PRD: zero
- Override rate on effort and confidence. A high rate means the proposals are off.
How I would evaluate the AI.
Grouping accuracy
Hand-label 50 feedback items, then measure how often the AI's themes match mine. Rerun after every prompt change.
Faithfulness
Every claim in a PRD must trace to an item id. Anything without evidence fails the check.
Calibration
Compare the AI's confidence with how often PMs actually accept the theme. High confidence with low acceptance means a prompt problem.
Checks already in the build
Each item sits in at most one theme, themes without valid evidence are dropped, and the PRD prompt forbids invented customers or numbers.
What could go wrong.
| Risk | How the product handles it |
|---|---|
| A wrong grouping looks confident | Evidence is always visible and any score input can be overruled. |
| Large accounts dominate the ranking | ARR weighting is shown openly. A later version adds a request-count view beside it. |
| Customer feedback is sensitive | Redact names before sending to a model, and keep an audit trail of what was sent. |
Where it goes from here.
Now
- Paste or upload a batch of feedback
- Export the ranked list
Next
- Connect the support inbox, call transcripts and CRM
- Let sales attach deal value when logging a request
Later
- Remember past decisions, including what was declined
- Track whether shipped themes saved the renewals they targeted
Shelfmate
A routine that builds itself, and a dashboard that proves it.
An AI routine builder for a skincare brand. Shoppers answer five questions and get a bundle built only from the store's catalog. The merchant gets conversion, order value and swap data to judge it. The skincare brand is invented.
Concept projectAI-nativeShopper and merchant viewsSimulated metrics
Tell us about your skin
Skin type
Main concern
Routine size
Budget
Your routine
Example for oily skin and breakoutsYou are the founder: set the bundle discount
Goals before you ship: revenue per session of at least ₹21, margin per session of at least ₹12, and a return rate under 5%. Drag the dial and watch the guardrails.
Simulated 8-week test, half of sessions with the routine builder and half without. Illustrative numbers.
Average order value by week
With routine builderWithout
Funnel, per 1,000 sessions
What shoppers change
Share of routines where the shopper swapped this product out.
Vitamin C and the bakuchiol serum are swapped most, mostly for price. A cheaper entry serum would likely lift bundle attach.
What shoppers typed in the free-text box
One in ten shoppers mentioned pregnancy or a skin condition. Those routines use gentle products and point to a dermatologist. The box is also free demand research for the brand.
Replace forty choices with one good answer.
A shopper lands on a skincare page with a wall of products and nobody to ask. The brand responds with discounts, which cut margin without helping the shopper decide. A short conversation that ends in a routine and a bundle helps the shopper first, and raises order value as a result.
Choice overload, from both sides.
Shopper
Too many options
They do not know which serum goes with which cleanser, so they leave or buy a single product.
Merchant
Discounts as the only lever
Small brands lean on offers to lift basket size, and margin pays for it.
Merchant
No proof for new tools
Every plugin promises more revenue. Without clean numbers, owners switch it off.
At Vudio.ai I worked with D2C brands every day and saw the third problem most. That is why the merchant view is half of this project.
A shopper and the person who runs the store.
Shopper · hypothetical persona
Ananya, 27
Combination skin, new to actives.
- Wants: a routine she can trust in under a minute.
- Fear: buying something that irritates her skin.
Merchant · hypothetical persona
Rohan, founder
Runs a ten-person skincare brand on a store platform.
- Wants: higher order value without more discounts.
- Fear: a tool that adds returns or complaints.
Five taps to a cart-ready routine.
- LandShopper opens a product page and sees the routine builder prompt.
- AnswerSkin type, concern, routine size, budget and an optional note.
- BuildThe AI picks one product per step from the catalog, with a reason.
- CheckRules verify catalog, steps, budget and safety before it shows.
- BuyOne bundle price, with swaps allowed, then add to cart.
Treated as a requirement, not an afterthought.
If a shopper writes that they are pregnant, allergic, on a prescription or have a skin condition, the routine is limited to gentle products and says to check with a dermatologist. The rule runs in the page and in the prompt. The free-text note is also treated as information about the shopper and never as instructions to the model.
Four checks on every routine
- Every product comes from the catalog
- One product per step
- Within the stated budget
- Gentle products only when a risk is flagged
If a check fails
The page falls back to a simple rules engine, so the shopper still gets a valid routine. They never see an error or an unsafe pick.
What I chose, and what I turned down.
A short form that ends in a routine
Shoppers on a product page want to tap, not type. Structured questions finish in under a minute and produce a cart-ready bundle.
Turned down: an open chat. It is slower, and the output is harder to turn into a cart.
The AI can only pick from the catalog
Every product id in the answer is validated. The model is a stylist working from this store's shelves.
Turned down: free recommendations from general knowledge, which could point to products the store does not sell.
The bundle discount is small and fixed
10 percent for three steps and 15 percent for four. The bundle is the better answer, and the discount only nudges it.
Turned down: deep, personalised discounts that train shoppers to wait for offers.
The merchant view ships on day one
Conversion, order value, funnel and swaps sit beside the shopper experience, because owners keep tools they can measure.
Turned down: shipping the shopper feature first and adding reporting later.
Revenue per session, with a seatbelt.
North star
Revenue per session, routine builder against control
From a 50/50 split of sessions over eight weeks.
Inputs
- Quiz completion rate
- Routine added to cart
- Bundle attach rate
- Swap rate per product
Guardrails
- Return rate does not rise
- Complaint rate does not rise
- Every flagged-risk session shows gentle picks and a professional note
How I would test the model.
A profile test set
About 40 shopper profiles, including sensitive skin, pregnancy, tiny budgets and contradictory answers. Check validity, steps, budget and the safety note.
Expert review
A dermatologist advisor reads a sample of routines each month and flags anything unwise.
What could go wrong.
| Risk | How the product handles it |
|---|---|
| Shoppers over-trust the routine | Plain wording, no medical claims, and a professional note when a risk is flagged. |
| A bigger basket brings more returns | Return rate is a guardrail in the experiment, and the test stops if it rises. |
| Catalog gaps give a weak routine | The swap data shows which steps shoppers reject, which tells the brand what to stock. |
Where it goes from here.
Now
- Quiz, routine, bundle price, cart
- Merchant dashboard
Next
- Real catalog and stock from the store platform
- A fragrance-free filter in the quiz
Later
- Refill reminders when a product runs low
- The routine builder on the cart page for single-product shoppers
I am looking for an APM or PM role in Bengaluru, in-office or remote. If your team is hiring, I would love to talk.