Case Studies
Designing human-controlled AI coaching for software engineers
Aude.ai
2025
Aude.ai told engineers where they fell short and did nothing to help them improve, so user adoption dropped by 60%. I led the design of a coaching feature from zero to one. Comparative testing with 27 engineers moved the product from an automation-first concept to a reviewable human-in-the-loop workflow.
Product Designer
3 Designers and 2 Stakeholder
MVP, User Research, Testing
40% Increase in feature adoption, 70% CSAT increase
UI Prototype - Cursor AI
Overview
Background and Role
Problem
The aude.ai platform diagnosed people. It never gave them direction to grow.
Research
Four findings, from 20+ interviews, a platform audit, and a competitive analysis
Pivot
My initial AI-driven concept was not implemented after pressure testing with users revealed two key concerns.
Designs
Where coaching sits in the product flow
Lessons
What I learned through the whole project

Making Booking Outcomes Clear: Reducing Support Queries by 25.8%
MakeMyTrip
2024
I redesigned Goibibo’s ground transport Thank You flow into a clear, state-aware experience that helped users understand successful, pending, and failed bookings, take the right next action, and discover relevant post-booking services without adding visual clutter.
Product Designer
Me + Project Manager
Android, iOS and mobile web for trains, buses and cabs; desktop web for trains
25.8% reduction in booking-status support contact rate.
Overview
Problem
The main problem was not the number of booking states. It was the way information was organized.
Research
Content Hierarchy Decisions
State-specific hierarchy
Designs
Design Decisions based on Use cases
Impact Metric
Primary KPI: Booking-status support queries and Secondary Metric: Cross-sell engagement

Multi Step Broker 0->1 CRM Workflow Management
Accolade
2024
Accolade had market data and an existing renter facing app, but no product direction for the broker side of the business. I ran research across 34 interviews and 8 competitor platforms, and it pointed away from our original hypothesis. The bottleneck was not property discovery. It was the broker, who sits at the center of every transaction and runs the entire deal on WhatsApp, phone calls, spreadsheets and memory. I designed a broker first CRM MVP that turns that scattered work into one operational system. It reduced lead management steps by 68 percent and saved brokers roughly 6 hours a week.
Product Designer
Aug 2024 - Dec 2024
3 Stakeholder and 3 Designer
Real Estate · SaaS · MVP Design ·
UI Prototype Coded Link
Overview
Context
Research
Research Revealed That Brokers Were the Operational Center of the System
Pivot
AI Preference matching model with brokers that did not worked out
System
System Architecture
Designs
User flow and Decisions
Impact
Brokers were able to complete core tasks reviewing leads, understanding client preferences, and tracking progress with significantly fewer steps.

Merchant Statements Email Design for Faster Financial Analysis
GradientFI
2025
Turning a static monthly report into a tool cross border merchants use to make decisions
Product Designer
May 2026
Solo Designer, End-to-End
Cross border, blockchain based payments
