Focused application · Agoda CEG · Product Designer L2/Senior

I’m a Senior Product Designer. I turn ambiguous, systems-heavy problems into products that ship and scale.

5 years across BrowserStack, Airlearn and Frim, from 0→1 product bets to consumer growth and enterprise workflows. This page is the short version; full case studies live in the portfolio.

0→1 product systems Consumer growth & retention AI-native interaction Enterprise workflows
Selected work

Three recent problems, compressed.

Each tab shows the problem and impact. Open one for decisions, evidence and what I learned.

AirlearnFounding Product Designer · 2024

How do you get users past the moment they usually quit?

Problem

Free users stopped after 2–3 lessons per session. Users who crossed 4 were 3× more likely to return the next day.

Move

Ran behavioural analysis, then redesigned the session loop around XP, ranks, contextual nudges and reward progression.

2.9 → 4.3average lessons completed / session
28% → 39%Day-1 retention
Learning: The retention problem wasn’t content quality. It was helping users cross the behavioural threshold where one more lesson became worth doing.
Read full case study ↗
Mondee / AarnaProduct Designer · 2025

How do you make invisible AI states understandable under engineering constraints?

Problem

Voice AI had four states users couldn’t see. Rive was new and risky for engineering.

Move

Designed the full state system plus a lightweight motion fallback, so clarity didn't depend on Rive shipping.

5 dayscomplete flow designed, animated and handed off
3 of 5usability participants chose voice over touch
Learning: Good interaction design should survive technical compromise. The state model mattered more than the animation technology.
Read full case study ↗
BrowserStack / App LCASenior Product Designer · 2025–26

How do you prove setup friction is blocking users from reaching value?

Problem

97 customer groups entered onboarding over ~60 days. 12 reached the first AHA (first DSL execution). The biggest break: a 49% drop at app upload / App Live dependency.

Move

Defined 3 AHA moments, shipped sample apps as the quick fix, then drove an AI-first onboarding and authoring redesign.

97 → 12groups entering funnel → first AHA
49% droplargest single-step loss
~2 mo → ~2 wkseparate customer-blocking P0 shipped
Learning: Users were being asked to complete infrastructure setup before experiencing product value. The redesign focuses on reducing pre-value setup and getting users to an executable test earlier.
$30Kenterprise deal unblocked via Step Settings
21 users / 14 daysStep Settings adoption with zero onboarding
~2 months → ~2 weeksDDT from problem to production
See BrowserStack work ↗
Founder · Product Designer · Builder

Frim taught me to own the product, not just the pixels.

I framed the problem, decided what not to build, designed it, shipped it, talked to users and changed direction when the data said to.

The three projects above show how I work inside teams. Frim shows what happens when there's no PM, no roadmap and every product call is mine.

I built a consumer travel app from zero: problem, design, ship, learn, pivot.

01Started broad · AI trip planning, itineraries, travel utilities
02Learned fast · More features made the product harder to explain
03Cut scope · Simplified to one job: preserve the story behind a souvenir
Why this matters for Agoda

CEG needs designers who can make product calls under ambiguity and simplify messy systems into clear traveler experiences. Frim is proof I can do that.

240 users · no paid push 700+ Android 130+ iOS
GTM experiment · Sri Lanka

I printed QR flyers, placed them in 20+ hostels, cafes and tuk-tuks across Sri Lanka. Zero ad spend.

124flyer scans
44beta signups
35%scan → signup
5partnerships

Every Sri Lanka signup came from physical placement. Every India signup came from LinkedIn. Context-driven acquisition beats audience-driven content for a travel product.

Read the full GTM note ↗
Agoda-specific product exercise
CEG · post-booking · concept exercise

What happens when “Confirmed” stops being true?

I traced one repeated post-booking failure: a confirmed stay becomes unavailable, and the traveler ends up coordinating the recovery.

Illustrative concept · grounded in public Agoda policy + qualitative complaint research
  • Public sources only · directional research, not incident-frequency data.
  • Competitor benchmark: Trip.com Room Guarantee.

When a stay breaks, coordination shifts to the traveler.

I checked customer complaints, Agoda's own recovery policies, and how Trip.com makes the same promise visible.

How I got here

Customer signal → repeated unavailable/mismatched stays.   System check → Agoda already has relocation rules.   Benchmark → Trip.com makes recovery explicit.

Harsh's real upcoming Agoda booking for Furama RiverFront in Singapore
My actual upcoming booking.
Singapore · 30 Sep–5 Oct 2026 · 2 adults.
Capability
Recovery already exists operationally.

Agoda’s policy expects relocation, not only cancellation.

Behavior
Travelers become coordinators.

Complaints show repeated Agoda ↔ property follow-up.

Benchmark
Trip.com makes the promise visible.

Recovery rules are surfaced before support is needed.

InsightAgoda has the recovery machinery. The opportunity is to make ownership and next steps clear to the traveler.

Three jobs when a confirmed stay breaks.

CertaintyIs my trip still safe, and what happens next?
RecoveryFind an equivalent stay without restarting search.
ProtectionShow the price, policy and trade-offs before I choose.
Design challengeHow might Agoda own recovery without making the traveler coordinate it?
DisruptionUnderstandCompareChooseResolved

1. Own the outcome, not the handoff.

2. Preserve the original booking intent.

3. Ask the traveler for one clear decision.

Put recovery inside the trip, not beside it.

Start in My Trips, open a focused choice only when needed, then return to the updated itinerary.

9:41● ● ●
⌖ Map•••
Singapore
30 Sep – 5 Oct 2026 · 🧳 1 booking
Itinerary
Saved
Day 1 · Wed, 30 Sep
Confirmed
Furama RiverFront30 Sep 14:00 – 5 Oct 12:00
Singapore
1 × Superior2 Adults · Harsh
✦ Suggested
Things to do

Most popular in Singapore
Need a flight?

Mumbai → Singapore
⌖ Map•••
Singapore
30 Sep – 5 Oct 2026 · 🧳 1 booking
Itinerary
Saved
Action needed · stay recoveryFurama RiverFront can no longer honor this booking.
!

Your stay is protected. We’re finding comparable options.

Issue foundOptions readyYour choice
Day 1 · Wed, 30 Sep
Recovery in progress
Furama RiverFront30 Sep 14:00 – 5 Oct 12:00
Original booking
1 × Superior2 Adults · Harsh
✦ Suggested
Things to do

Most popular in Singapore
Need a flight?

Mumbai → Singapore
Stay recoverySingapore · 30 Sep – 5 Oct
Issue
2Compare
3Confirm
Original bookingFurama RiverFrontSuperior · 2 adults · 5 nights
Closest match

Choose a replacement without searching again.

Matched to your dates, guests, area, room and terms.

2 other comparable optionsCompare →
⌖ Map•••
Singapore
30 Sep – 5 Oct 2026 · 🧳 1 booking
Itinerary
Saved
Stay recoveredYour Singapore stay is secure.
✓ Replacement booking confirmed✓ Original booking closed✓ Updated voucher ready
Day 1 · Wed, 30 Sep
Confirmed
Comparable stay · Singapore River30 Sep 14:00 – 5 Oct 12:00
Replacement booking
1 × Same/better room2 Adults · Harsh
✦ Suggested
Things to do

Most popular in Singapore
Need a flight?

Mumbai → Singapore

Keep the traveler flow simple; orchestrate complexity underneath.

Route each case to auto-resolution, a traveler choice, or an agent.

DetectProperty cannot fulfilRejection, overbooking or inventory mismatch.
ValidateConfirm state + urgencyUrgency, payment and supplier state.
UnderstandCapture original intentRoom, guests, location, terms and price.
SearchFind viable inventorySame property, then comparable nearby stays.
Self-serveEquivalent option + clear policy + holdable inventory.
OR
AgentNo viable option, unclear liability or payment exception.
RankScore replacement equivalenceRank against the original booking.
ProtectCalculate cost + policyTraveler cost + recovery policy.
SwapHold new → close oldHold replacement before closing original.
ResolveConfirm + preserve contextConfirm, update voucher, preserve context.
Traveler sees
ProblemProtectionOptionsOne decisionResolved
Agoda coordinates
PropertyInventoryPolicyPaymentBooking stateSupport

Measure recovery, not support activity.

PrimarySuccessful recovery rateDid the traveler accept a replacement?
CustomerTime to certaintyHow quickly did they get a viable path forward?
EfficiencySupport contacts / disrupted stayDid recovery reduce repeat contact?
QualityReplacement equivalenceHow closely did the replacement match the original?
GuardrailRecovery rejection / secondary failureDid the replacement fail or get rejected?
Product betMake Agoda own the outcome until the stay is secure again.