I build design culture, systems, and strategy — and stay hands-on from first sketch to shipped product. From founding designer at RedBus to heading design at a unicorn, I've shipped across fintech, telecom, agritech, and AI in India and the US.
Designed an intent-driven conversational commerce experience for a US MVNO using the ChatGPT SDK — a 3-layer architecture mapping SDK components to a custom brand system, with 4 conversation flows and full edge case coverage.
End-to-end redesign of Perfios's RBI-regulated Account Aggregator consent app. Uncovered a 13.54% dead-click rate and 20–25% drop-off through rigorous research, then redesigned around user trust and data transparency for 200K+ users.
Designed a first-of-its-kind socially-conscious US MVNO where buying a mobile plan triggers a free data+call+text connection for someone in the developing world. Travelled to Ranchi to observe recipients firsthand — research that reshaped the entire impact experience.
Enter the password to view the selected case studies.
I believe the best design disappears — leaving only clarity, trust, and a person who accomplished what they came to do.
My career has been defined by context diversity: I've designed for rural farmers who can't read, regulated fintech users who are afraid to share their data, US millennials choosing mobile plans, and bus travelers booking across India in a dozen languages.
Each of these contexts demanded field research, not assumptions. Systems thinking, not single screens. Design that earns trust rather than demands it.
Along the way I've built and led product design teams of 4, 6, and 10 — hiring designers, mentoring interns, and standing up design practices where none existed before.
I'm based in Bangalore, open to relocate, and looking for a design leadership role where craft and strategy are equally valued.
Returned to Reach to lead · AI conversational UX · Design system from scratch
2023–24
Head of Product Design
Perfios.ai
Unicorn · RBI-regulated fintech · Anumati redesign · Built & led a 10-person design team
2021–23
Lead Product Designer
nurture.farm
UPL Subsidiary · Rural India · Low-literacy design · Built a 6-person team & mentored interns
2018–21
Lead Product Designer
Datami / Reach Mobile
US MVNO · 0→1 · Gigs social impact feature · Built a 4-person product design team
2017–18
Lead Product Designer
Bizom
B2B SaaS · Paper-to-scan insight · Voice ordering
2011–16
First Designer
RedBus.in
20M+ users · B2C + B2B · Foundational UX still live today
03
How I Work
01
Field first, assumptions never
Every significant decision I've made was shaped by watching real people use products in their real environment — rural India, Ranchi, retail shops across the country. The insight is always in the gap between what people say and what they do.
02
Trust is the hardest problem
Designing for fintech, healthcare, or any regulated domain is fundamentally about trust. Users who don't trust a product won't use it — no matter how elegant it looks. I design for transparency, clarity, and control above all else.
03
Systems outlast screens
Single screens don't scale. The real work is building design languages, component systems, and team rituals that sustain quality long after the initial launch. Every product I've led has a system underneath it.
04
Design is a business function
I measure my work in outcomes: activation rates, drop-off reduction, adoption, revenue. Beautiful work that doesn't move a metric is decoration. I sit in product and business conversations because that's where design decisions are really made.
05
Context is the brief
A farmer with an intermittent 2G connection and a US millennial choosing a phone plan have radically different needs. Generic solutions serve nobody. I've spent my career designing for specific, difficult contexts — and that specificity is what makes products work.
06
Curiosity is a leadership skill
The best design teams I've been part of were relentlessly curious — about users, about technology, about the world. I write publicly about design, AI, and judgment. Learning in the open makes the whole team sharper.
04
Expertise
◈
Design Leadership & Team Building
Building design culture, rituals, and review processes that sustain quality at scale
◈
0→1 Product Strategy
From blank canvas to shipped product — ecosystem analysis, value proposition design, rapid validation
◈
Design Systems at Scale
Cross-platform systems with token architecture, component libraries, and documentation
◈
Conversational & AI UX
Intent-driven flows, ChatGPT SDK integration, trust design for AI-assisted decisions
◈
User Research & Synthesis
Field research, FIU interviews, usability testing, affinity mapping, behavioural analysis
◈
Regulated Domain Design
RBI-compliant fintech, data privacy, consent flows — balancing legal rigour with usability
◈
Cross-functional Collaboration
Embedded in PM, engineering, and business teams — speaking the language of each
◈
Accessibility & Inclusion
Designing for low-literacy, multi-language, low-connectivity, and first-time digital users
◈
Stakeholder Management
Presenting design decisions to executives, regulators, and cross-functional leads with clarity
05
Writing
On Medium, writing about AI, conversational interfaces, and the design judgment that doesn't show up on screens.
I'm looking for a design leadership role in India where craft, research, and business impact are equally valued. If that sounds like your team, I'd love to talk.
Designing an intent-driven AI commerce experience for a US telecom brand — where the conversation is the product.
Company
Reach Platform
Role
Head of Product Design
Technology
ChatGPT SDK
Scope
0→1 · Full System
The Problem
Telecom plan selection is broken by design
Choosing a mobile plan is one of the most frustrating consumer experiences in the US. Dozens of plans, jargon-heavy comparisons, hidden fees, and rigid decision trees that assume everyone wants the same thing. Users drop off. Confidence is low. Regret is common.
At Reach Platform — a US MVNO with a social mission — we asked a different question: what if instead of filtering options, we just listened to what people actually needed?
The Insight
People already know what they want — they just can't find it
Users describe their needs in natural language: "I travel a lot", "I need something cheap for my kid", "I want to know if my phone works before I switch." The existing UI couldn't listen. We could build one that did.
"This is an intent-driven conversational state machine — not a fixed linear flow."
The key design insight was that a conversational experience for commerce is fundamentally different from a chatbot. It needs to handle mid-conversation changes, remember context across turns, enforce eligibility guardrails, and still complete a purchase. Those are design problems, not engineering ones.
The Architecture
Three layers. One seamless experience.
The core design challenge was: how do you build a brand-grade commerce experience on top of an AI SDK that controls the chat framework? We designed a 3-layer component architecture:
01ChatGPT SDK Layer — chat bubbles, chips, input, layout. Provided by the SDK, customised for tone and brand colour.
02Commerce & Domain Layer — Plan Cards, Device Cards, Bundle Cards, Coverage Tiles, Recap Strip, Checkout CTA. All built from scratch in the Reach design system.
03Brand Layer — color tokens, gradient motion, typography, brand animation easing. SDK design tokens mapped 1:1 to Reach's token system.
The Superflow — Intent-driven, Non-linear
User says anything
e.g. "Help me find a good mobile plan"
↓
01
Plans
Clarify needs → Show plan cards → Select
→
02
Devices
Optional → Fetch catalog → Select device
→
03
Protection
Device-aware → Upsell if device selected
→
04
SIM
eSIM or physical → Eligibility check
→
05
Checkout
Review → Confirm → Hand off to payment
Non-linear
User can jump between stages, change their mind, or ask off-topic questions at any point — the system holds context and resumes intelligently.
The Flows & Guardrails
Four scenarios. Eligibility rules. Dozens of edge cases.
We designed four primary conversation flows — Apply Same Plan to All Lines, Mix & Match, User Changes Mind Mid-way, and User Asks for Details — plus a full device selection sub-flow. Every edge case was documented: what happens when coverage fails, when a device is ineligible, when a user backtracks.
The guardrail system was equally critical. Unlike a fixed form, a conversational flow can go anywhere — the user might ask about phone compatibility mid-purchase, or jump from coverage straight to checkout. Guardrails defined eligibility rules that kept the experience coherent.
Design Principle
❌ Traditional UI
Step 1
→
Step 2
→
Step 3
→
Step 4
Fixed linear flow. User must follow the path.
✓ Conversational State Machine
Plans
⇄
Devices
⇄
Protection
⇄
Coverage
⇄
Checkout
Intent-driven. User can move between states. Context is preserved.
"This is an intent-driven conversational state machine with eligibility rules — not a fixed linear flow."
Guardrail Rules — What the AI Can & Cannot Do
Scenario
Eligibility Rule
Why It Matters
🛒
Checkout
Plan required
Can't purchase without a selected plan — prevents broken orders
🛡️
Device Protection
Device required
Protection is device-specific — only shown when a device is in cart
📡
Coverage Check
Allowed anytime
Users often want to verify coverage before committing — always available
📱
Buy Device
Plan required first
Device catalog is plan-dependent — shown only after plan selection
💬
Random Question
Answer → resume flow
AI handles off-topic gracefully, then steers back — never gets stuck
The Outcome
A commerce-grade AI experience shipped to production
The Reach conversational plan selector is one of the first production examples of a full ChatGPT SDK-powered commerce journey — from plan discovery through device selection, coverage check, protection options, SIM type, and checkout — in a single conversation thread. No page loads. No decision trees. Just a conversation that ends in a purchase.
The Flow — Screen by Screen
From "Help me find a good plan" to order complete
Every screen below is a real frame from the production experience — running inside ChatGPT.com via the Numobile connector.
Step 1 — Connect & Intent
OAuth-style trust moment — user approves Numobile to connect to ChatGPT
User types naturally: "Help me find me a good mobile plan" — AI begins clarifying
Step 2 — Clarify Needs (Conversational)
AI asks: "Just you, or family members?" — context-aware, not a dropdown
User responds: "Family plan, 3–4 lines" — AI continues naturally
AI asks for ZIP code to check local coverage — progressive data collection, not a form
Step 3 — Plan Selection (Custom Commerce Cards)
Custom Plan Cards surface inside the conversation — price, discount badge, features, Broadband Facts
Plan selected (✓ Selected state). AI asks: apply to all lines or mix & match?
User chooses "Same plan for all" — AI confirms: By the Gig (1GB) selected for all 4 lines. Moves to devices.
Step 4 — Device Selection (Skeleton → Loaded)
Skeleton loader state — shimmer cards while device catalog fetches. Designed to feel native, not like a web load.
Device Cards appear — iPhone 17 Pro, iPhone 16 Pro with pricing, discounts, and financing options inline
Step 5 — Protection, Review & Checkout
Device Protection Card — contextual upsell, device-specific pricing, discounted badge, "Added to Cart" confirmation state
Full order review inside the conversation — monthly charges, device charges, one-time fees. No surprises at payment.
Payment page on Numobile.com — card, Amazon Pay, Cash App Pay. Handed off cleanly from the conversation.
Step 6 — Order Complete
Order complete. "Welcome to NU Mobile! Your order is complete." — eSIM activation next. The full journey: one conversation.
Anumati — Consent Redesign · Perfios
Case Study 02 · 2023–24
Anumati — Consent Redesign
Redesigning a regulated financial consent experience for 200K+ users — where trust is the product and confusion is the enemy.
Company
Perfios.ai
Role
Head of Product Design
Users
200K+
Regulation
RBI AA Framework
The Problem
Users were losing trust before they'd even started
Anumati is Perfios's Account Aggregator consent app — the interface through which Indian consumers control how their financial data is shared between institutions, under the RBI's AA framework. It sounds important because it is. But the existing product was failing its users.
13.54%
of sessions had dead clicks — users tapping on elements that did nothing
20–25%
user drop-off between login and account discovery
20%
further drop-off from account linking to consent approval
1.7 min
average session — only 1.3 min active. Users were confused, not considering.
The Research
The core problem was trust, not usability
We ran FIU (Financial Information User) interviews, dead-click analysis, user interviews, and usability testing. The data pointed to a usability problem. The conversations revealed something deeper.
"Through my mobile number you got all my accounts, but I don't know what you are accessing."
Users weren't dropping off because they couldn't complete the flow. They were dropping off because they didn't understand what they were agreeing to, who was accessing their data, and whether they could take it back. The product was asking for trust it hadn't earned.
Other revealing moments from user conversations: "Do I have to link all the accounts?", "I think I'm done, but I'm re-directed to the previous page", "Was there something written about the data? I didn't read."
The Design Response
Make data control visible, granular, and reversible
01Simplified terminology — eliminated AA ecosystem jargon ("FIP", "FIU", "handle") and replaced with plain-language explanations of what was happening and why.
02Transparent data display — made it explicit which institutions were accessing what data, for how long, and under what terms — before asking for consent.
03Reduced OTP friction — the multi-OTP flow was causing abandonment. Worked within RBI constraints to streamline verification without compromising compliance.
04Recovery paths — designed clear states for "no accounts found" and error conditions, so users knew exactly what to do next rather than re-trying blindly.
05Native app feel — research showed users trusted the experience more when it felt like a native app. Applied brand-consistent components to reinforce familiarity and reduce drop-off.
Step 1 — FIU Interviews · Systemic Pain Points
The institutions were struggling too
We started with the people who operate the system — Financial Information Users (FIUs). Their frustrations revealed that the problems weren't just UX-level. They were systemic.
AA success rate is not very high — FIUs feel helpless to resolve this
When a major FIP is down, customers call the FIU for support
FIU keeps enabling/disabling Account Aggregator based on which bank the user selects
Onboarding FIU with deep customisations takes a lot of time
FIU don't want to cater to more use-cases until the ecosystem is more stable
KreditBee — 70% of users shown AA as option choose to proceed with it
HDFC — only 1/4th of customers are choosing AA
Money View — 40% users choose AA as mode of data sharing
→
Design implication: The experience needed to work for both ends — reduce FIU's operational overhead and increase user confidence enough to choose AA over alternatives.
Step 2 — Addressable Numbers · Quantified Failure
Data defined the brief, not assumptions
Dead-click analysis and behavioural tracking gave us hard numbers. These became the design brief — every decision was measured against reducing these specific metrics.
13.54%
Sessions with dead clicks
Users tapping elements that did nothing — invisible UI friction
Login → Account Discovery
20–25%
Drop-off at first step
Users leaving before seeing their accounts — before anything useful even happened
Account Linking → Consent
20%
Drop-off mid-journey
Users who made it through login still abandoning before consent approval
1.3 min active1.7 min total
1.7 min
Average session
24% of time was inactive — users stuck and confused, not deliberating
Step 3 — User Research Discoveries · Behavioural Insights
What users actually experienced
User interviews and usability testing revealed the behavioural patterns behind the numbers — confusion loops, trust gaps, and broken recovery states.
Users have difficulty understanding terminologies in the AA ecosystem
Account aggregator is not a default choice unless a benefit is specified
Users don't recognise or trust Anumati on its own
Users only scan the elaborate details — distracts from completing the main task
User has no idea why accounts are not shown after mobile OTP step
User has no idea how to recover from "no accounts found" — keeps re-trying
Users spending more time trying to understand account linking section
After completing journey, users unsure what data was shared, with whom, or for how long
Users feel they are repeating steps when going from accounts linking to consent approval
Users seem to trust the FIU/FIP more — and complete the journey through them
Step 4 — Candid Insights · Users in Their Own Words
The trust crisis, unfiltered
These aren't paraphrased. These are verbatim moments from user conversations that defined the redesign direction — trust, not usability, was the root problem.
"
Through my mobile number you got all my accounts, but I don't know what you are accessing
→ Redesigned consent screen to show exactly what data, with whom, for how long — before approval
"
Do I have to link all the accounts?
→ Added granular account selection with clear "verified" state and guidance tooltip
"
I would like to give only what is required
→ Made consent granular and user-controlled — reject consent always visible
"
I think I'm done, but I'm re-directed to the previous page
→ Added explicit progress bar and clear success/handoff state
"
Was there something written about the data? I didn't read
→ Replaced dense text with scannable consent summary — purpose, data type, duration at a glance
Step 5 — Design Heuristic Study · UI Diagnosis
10 UI violations. All addressable.
A heuristic evaluation of the existing interface catalogued the UI-layer problems systematically. Each finding became a specific design fix.
01
UI design is off putting — leading to higher cognitive load
→ Full visual redesign following brand and accessibility standards
02
Multi-consent lacks context for each consent
→ Consent cards now show purpose, data, and duration individually
03
Multiple expanded consents push main action out of view
→ Collapsed consent list with sticky CTA always in viewport
04
Skeuomorphism creates false clickable affordances across the page
→ Flat design system with clear interactive vs informational states
05
Improper content hierarchy — user actions feel unguided
Web UI not space efficient — unnecessary scrolling
→ Compact card layouts, removed decorative elements
07
Loader implementation not up to brand standards
→ On-brand shimmer loaders matching Anumati colour system
08
UI design not following typographic standards
→ Applied Anumati type scale consistently across all screens
09
Icons from different families — broken brand experience
→ Single icon library, tokens enforced in design system
10
Responsive and adaptive experiences are broken
→ Mobile-first redesign with breakpoint testing
Step 6 — What Are We Solving For · How Might We
Three problem spaces. Six design questions.
Research synthesis led to HMW questions that framed the redesign across three domains — each one traceable to a specific research finding.
Customer Experience
How might we make the account aggregator ecosystem familiar for users?
How might we make efficient integrations with FIUs?
User Experience
How might we save our users' time?
How might we clearly communicate with the users?
User Interface
How might we keep consistency across touch points?
How might we alleviate heavy cognitive load for the users?
Step 7 — Design Process · Double Diamond
Every decision traceable to a research finding
We ran the full double-diamond — not as a formality, but because the problem had four distinct layers. Each phase had specific outputs that fed the next.
Discover
FIU Interviews
Dead click analysis
User interviews
Define
Heuristic study
HMW reframes
Problem framing
Develop
User journey map
Low-fi wireframes
Usability test
Stakeholder pres.
Deliver
High fidelity mockup
Stakeholder sign-off
Usability validation
The Final Design — Single Consent Flow
From confusion to clarity — the redesigned journey
The redesigned Anumati consent flow makes data sharing transparent, granular, and user-controlled at every step. Here is the complete single-consent journey as shipped.
Entry — Trust established before data is requested
01
FIU Trigger
User is shown Account Aggregator as the recommended option — with clear benefit framing. No jargon.
→
02
Enter Mobile
Simple, single-purpose screen. "Is my data safe?" persistent link builds trust before OTP is requested.
Account Linking — Verified state with contextual guidance
03
Contextual Toast
Instead of silent success, a tooltip guides: "1 account added, add more or click Next." Eliminates the loop of users re-trying steps.
→
04
Verified / Not Verified
Clear separation of verified and unverified accounts. User knows exactly where they are and what to do next.
→
05
Account Selection
"Powered by RBI regulated Anumati" footer — trust signal at the exact moment of decision.
Consent — Full transparency before approval
06
Consent Detail
Purpose, data type, verified accounts — all visible before consent. Answers "what are you accessing?" before the user has to ask.
→
07
Approve Consent
Single, clear CTA. "Don't want to provide consent? Reject consent" — giving users agency, not hiding it.
→
08
Consent Approved
Clean success state. User is handed back to the FIU (Axis Bank) — the handoff is explicit, not abrupt.
Live Prototype — Consent Flow in Action
What to watch for
01
Progress bar at top
Persistent visual indicator — user always knows where they are in the journey
02
"Is my data safe?" link
Trust anchor available at every step — not buried in settings
03
Contextual tooltip on linking
Guides user after each account is added — eliminates the "am I done?" confusion
04
Consent detail before approval
Purpose, data type, and duration shown before the approve button — trust earned, not assumed
05
Clean handoff
"Consent Approved — Moving back to Axis Bank" — explicit, not abrupt
🎯
The result
Congratulations Abhilash — Rs. 4,00,000 instant disbursement approved. The consent flow that was losing 40–45% of users now completes seamlessly. Trust earned, data shared, loan approved.
Gigs — Social Impact MVNO · Datami / Reach Mobile
Case Study 03 · 2018–21
Gigs — Social Impact MVNO
Designing a socially-conscious mobile service from zero — where buying a plan gives free connectivity to someone who needs it most. 3-person team · 4 months · US market.
Company
Datami / Reach Mobile
Team
3 designers · 4 months
Market
US Millennials
Type
0→1 · MVNO · Social Impact
Challenges and triumphs while designing a socially-conscious mobile service
The Brief
A bold starting point
"We need to build a digital-driven mobile service (no offline store) for millennials in the United States"
?Why we need to address this?
?Who are our potential customers?
?What is the current behaviour?
?Frustrations?
?Expectations?
?How can we kick start remotely?
My Contributions — 6-Stage Process
Research → Strategy → Scenarios → Structure → Surface → Final
01
Research
→
02
Design Strategy
→
03
User Scenarios
→
04
Structure
→
05
Surface
→
06
Final
Stage 1 — Research
Multi-method research before a single screen
The marketing and business team ran market research in parallel. On the design side, we ran focus group surveys → 1-on-1 interviews → synthesis via affinity diagram — all before opening Sketch.
❌ Statement of Limitation
Current providers follow the legacy of offline store experience
Hidden charges and complex tax calculations
Users pay the full plan price irrespective of data usage
No delightful elements — purely transactional
✓ Statement of Vision
Convenient online experience — say goodbye to offline stores
No hidden charges, simple and transparent billing
Pay for what you used
Make a difference. Get a plan, give a plan.
Stage 2 — Design Strategy
User persona × brand values
Four user archetypes mapped across Convenience ↔ Feel-good and Individual ↔ Group axes. Brand partners like Airbnb, TOMS, and Warby Parker informed the positioning — brands that make people feel good about spending.
Stage 3 — User Scenarios & Tasks
Context story + functional tool = functional elements
Purchase
Overall value proposition
Sign up / log in
Plan details
Billing & shipping info
Track order
Activation
Barcode scanning
Activation flow
Account information
Billing details
Invoice
Engagement
Data usage
Impact story + video
My Impact
Push notification
In-app notification
Stage 4 — Structure · UX KPIs
Measuring what matters
01
Time to finish the task
02
Session rate
03
Major drop-offs
04
Feedbacks (Loved it, Happy & Unhappy)
Medium-fidelity wireframes tested the core tasks: plan selection with social good framing, IMEI validation (a key drop-off trigger), and the activation flow.
Stage 5 — Surface · Brand & Visual Design
Human. Simple. Approachable. Local.
Collaborated with brand identity designer Jennah on the visual language. The final Reach Mobile mark — chosen from 6 directions — captures warmth and approachability without feeling corporate.
Human
Simple
Approachable
Local
Stage 6 — Final Design · Shipped Product
8 screens. Every pixel considered.
The complete Reach Mobile app — from discovery through activation and engagement — designed end-to-end in 4 months with a 3-person team.
01
Home
Personal impact stats, story cards, and referral — all above the fold. "Good morning, Abhilash" — personalised from day one.
02
Plan Details
Sharable · Flexible · Giving. Every plan feature tied to a value, not a spec. "We donate 1GB to a family in need" — the cause built into the product copy.
03
Account — Empty State
Illustrated empty state instead of a blank screen. "Missing my account?" — warm, human tone even before the user has signed in.
04
Welcome Aboard
Post-signup state showing all active lines. "Welcome aboard, Abhilash!" — celebratory but functional. Each line named and status shown at a glance.
05
My Impact
30 days of connectivity given. 102 recipients. 300 days. The impact shown in human terms — with recipient photos and #GigsThatGive shareability.
06
Add Lines
"Sharing is caring" — add up to 3 lines to share your data plan. Simple line management with a single tap to add, $30/month per line, clear hierarchy.
07
Scan Barcode
SIM activation via barcode scan — eliminating manual IMEI entry for the primary activation path. The Reach Welcome Kit has a barcode on the back of the SIM.
08
Device Compatibility
IMEI jargon was the #1 drop-off trigger in usability testing. This screen demystifies it — "How to find IMEI number?" with Android/iOS guides and a video walkthrough.
Post-Launch Iteration 01 — Referral
The referral wasn't getting traction. We fixed it.
Customer success data showed users were happy with the product and willing to invite friends — but the referral carousel wasn't triggering the action. Google Analytics confirmed low click and session rates on the referral module.
The Problem
The referral CTA was buried in a carousel — users were missing the trigger entirely. The referral credit wasn't visible enough to motivate action.
Design Iteration
Moved from a carousel to persistent in-context communication. Added referral credit to the account section as an always-visible anchor. Result: engagement improved significantly.
Before
→
After
Post-Launch Iteration 02 — Impact
Numbers aren't inspiring. Stories are.
Active users told us they liked the contribution numbers — but the numbers weren't connecting emotionally. They liked knowing they'd helped, but felt distant from the impact.
✈️
The Ranchi trip
I travelled to Ranchi for two days with a writer to understand how recipients actually lived with free connectivity — how they used it for work, family calls, their children's education. We needed to feel what we were designing for, not just imagine it.
"People need to know the story — not just the numbers."
The Problem
Impact was displayed as abstract statistics: 109 Recipients · 6.5M Min Voice. Users couldn't connect to the people behind the numbers.
Design Iteration
Reworked 'My Impact' to lead with stories and faces — real recipients, real names, real narratives. Numbers moved to a secondary position. Added "Share Impact Story" as a one-tap action.
