Interaction design, design systems and information architecture for mobile apps and platforms. More than ten years, including seven at Amazon.
Selected work
SensAiry
The wheel-position model drivers already understand, rebuilt with visible states, feedback and recovery.
Read case study
Fact Finders Pro
Trust patterns and a 50+ component system, from a lean prototype to a public launch.
Read case study
Smartflo
The brief said dashboard. Customer visits pointed at the navigation, and nine sections became six.
Read case study
Amazon
Brand stores, campaigns across Fire TV, tablets and display, and the standards that kept many designers on one bar.
Read case study
How a system becomes a product
This is the order I build in. The tokens and labels here come from my Fact Finders Pro design system, and the last frame is the real dashboard. Step through each stage, or press Replay to watch it build.
Good product foundations span interaction, components, structure and the people who use them. This is where each case study gives the clearest evidence.
| Case study | Native and mobile interaction | Patterns and components | Navigation and mental models | Standards and teaching |
|---|---|---|---|---|
| SensAiry | Drag-to-pair on iOS, redesigned with visible states and recovery | — | Every screen organized by wheel position | — |
| Fact Finders Pro | — | Tokens, type scale, input states and components with properties | Four user types narrowed to two | Mentored junior designers on the team |
| Smartflo | — | Welcome screen, setup checklist and empty states | Nine top-level sections regrouped into six | Design principles co-created for the business unit |
| Amazon | Store layouts designed one column first; Fire TV built for the remote | A tile system that absorbed three platform changes | — | A review framework, and onboarding adopted in two design hubs |
means the project isn’t the evidence for that area.
Four steps I follow on most projects, in this order. Each case study shows one of them most clearly.
I check the request against customers and data before designing the screen I was asked for.
SmartfloI structure the product around how people already think about the task, then make the system fit.
SensAiryWhen a problem repeats, I solve it once as a component or pattern that every new screen inherits.
Fact Finders ProShared criteria and onboarding let a whole team hold one standard without me in the room.
AmazonRecommendations from a manager and colleagues.
“In a span of 90 days demonstrated her grip on her subject by mining right insights to define action plan to co-create design principles and UI/UX framework for the business unit.”
“Her creative problem-solving and innovative thinking significantly raised the creative bar within the team.”
“She consistently comes up with innovative ideas and approaches, crafting elegant solutions for complex business problems.”
Work / SensAiry
Case study 01, mobile redesign
SensAiry pairs small sensors on each tire with the driver’s phone. I designed the original app in 2015. By 2026 it looked and behaved like 2015, and the Tymtix team asked me how to revamp it. My redesign keeps the one idea that still worked, the way drivers already think about their car, and rebuilds everything around it.
Problem
The 2015 app had aged: dense screens, decorative gauges, little feedback, and a setup flow that could trap you.
My role
Led the redesign end to end: informal research, information architecture, interaction and visual design, prototype.
Key decision
Keep the wheel-position model drivers already understand, and rebuild states, feedback and recovery around it.
Result
A modern, accessible redesign and a working prototype, shared with the team as the direction for the revamp.
Why redesign
The 2015 app was built for a 4-inch iPhone and the visual trends of its time. Ten years later it fell short of today’s basics: clear hierarchy, readable contrast, comfortable touch targets, and feedback for every state. One thing was still right: drivers say “front left,” not a serial number, so every screen is laid out like the car.
Before and after
The structure stayed recognizable, so existing users wouldn’t have to relearn the app. What changed is hierarchy, feedback and the details that make a task feel easy.
Research
In 2025, alongside the Google UX Design course, I used SensAiry as my subject and spoke informally with drivers from my own circle (names changed). Small and informal, but it shaped the redesign.
“I wish my car just told me when and how much to inflate.”
“I need a dashboard for all 25 vehicles. Simple.”
Both needs show up in the redesign: status and alerts sit on each vehicle tile, so a fleet reads at a glance, and alert thresholds are something people can set.
The same people grouped the app’s features on cards. Their groups became seven top-level areas.
Redrawn from my original information architecture.
The core interaction
Instead of typing a 12-digit serial for each tire, the driver drags each detected sensor onto the wheel it’s mounted on, which mirrors what they’re doing outside the car. The redesign makes every step of that visible and reversible.
Decisions and trade-offs
Visual system
Dark surfaces for use in a car or garage, large type, and color reserved for status, so the one thing that matters stands out. The same rules apply from the empty state to the fleet view.
How I used AI
Claude Code helped generate token names and variable structures and draft usage and accessibility notes for me to edit. Figma MCP read the file’s components and variables directly and flagged mismatched styles. The tools handled the bookkeeping of a system, so I could spend my time on the decisions that need taste.
The redesign and prototype went to the Tymtix team as the direction for the revamp.
A former colleague reached out informally, ten years after I designed the original, to ask how the app should evolve. Redesigning my own work was a useful test: the mental model held up, and everything around it was ready for a decade of better craft.
In 2015 I designed the app and its brand: the icon, the launch banners and a series of illustrated safety tips.
Work / Fact Finders Pro
Case study 02, patterns and design system
Fact Finders Pro helps investors evaluate startups with AI. A finding is only useful if an investor can see where it came from. I took the product from a lean prototype to a focused platform, designing its trust patterns and the component system that carries them.
Problem
Investors can’t act on AI findings they can’t check, and re-checking by hand erases the time AI saves.
My role
Sole designer from the first prototype through the redesign; later, design contributions to Marketplace and reporting.
Key decision
Define trust once as patterns every AI output inherits, and focus on two core audiences instead of four.
Result
A 2025 lean prototype became FFP Smart, launched publicly in June 2026. I continue to review builds and design new flows.
I owned
Others owned
Problem
AI can read a startup’s data room quickly. But if an investor has to verify every AI finding by hand, the time saved disappears. So the design problem wasn’t generating analysis. It was making each finding checkable in place, without leaving the screen. There were no written specs and direction changed often, so the design file became the spec.
Evolution
The design was iterated through internal task walkthroughs and feedback from product and engineering. Three stages show how the product found its shape.
2025: the lean prototype. One form, one upload, one result. It proved the core loop before anything else was built.
A first visual style, explored and dropped. A bold gradient onboarding. When the strategy shifted, rebuilding cost less than patching it.
The prototype served investors, startups, analysts and admins on the same screens. I narrowed it to two primary users: investors, who assess startups and verify facts, and startups seeking investment.
Trade-off: analysts work in a view designed for investors.
When the strategy shifted, I dropped the first visual direction instead of adapting it, and rebuilt around verification.
The pattern library
Rather than designing trust feature by feature, I defined a small set of patterns that any new screen inherits. A score, a claim or a chart all answer the same questions the same way.
An evaluation end to end: checking the data extracted from a startup’s documents, the data room and its questionnaire, then moving between modules and opening the evidence behind each finding in a side panel. Recorded from the product and sped up.
Designing the wait
Generating a diligence report isn’t instant. I designed the report experience with engineering, aligned with how the product already behaved: confirm what will be generated, show that it’s in progress, say when it’s ready, and keep every past version.
Reports cost credits and depend on a plan, a payment and complete data, so most of the design work went into what happens when something isn’t right. Every message says what happened, what it cost, and what to do next. The wording below is from the designs.
Three rules held across every state
In motion
The in-progress state uses a quiet looping indicator and plain language, “Request accepted, in progress,” instead of a spinner with no explanation.
Foundations
More than 50 components on one set of tokens, handed to engineering as the source of truth. These specimens are rendered from the values in my FFP style guide and Figma tokens.
Display XL shown reduced to fit.
Gradient CTA
Action Card
iconTitleDescriptionButton instanceCompany card
Anatomy from the FFP Figma library; property names are the real ones. Even the loader is a component set: 4 animation states × light and dark backgrounds, 8 variants.
Type, color and components carry from the product to the marketing site, so an investor meets one consistent brand before signing in.
Process
Direction came through conversations and AI-generated prototypes, not PRDs. So flows were written as scenarios and rules next to the screens they shaped, and every flow was wired as a clickable prototype.
Prototype wiring
Every screen and pop-up was connected, so internal walkthroughs used real flows rather than static screens.
A planning tool I built from the product meetings I attended, not from user research. For each step of adding a startup, it pairs what an investor is thinking with what the product should do. Those responses became the Add Startup design: few fields, autofill where possible, clear progress and an obvious next step.
I want to get this into my pipeline quickly.Clear entry points; a fast, uncluttered dashboard.
This should be fast. I’ll use this often.A prominent, consistent Add Startup action.
Let me get this started.Several entry points: dashboard, data room, search.
I hope this is quick. Can it autofill?Autofill from a URL; only the essential fields.
Good. Now I can start reviewing.Confirm it clearly; offer the next steps.
Redrawn from my original journey map.
A rule, not just a screen
The repeat-user sign-in flow carried a product rule as a note: allow five attempts, then lock the account and point the user to the forgot-password flow, rather than adding administrative overhead.
Trial account rules, as written in the file
A trial allows 3 startup evaluations and lasts 30 days. No payment method is collected during the trial; it’s collected on upgrade. 15 days before expiry, the investor gets a reminder to upgrade, and the account then enters a grace period.
Paraphrased from the FFP MVP Figma file
An admin scenario, before the internal portal was wireframed
A super admin adds a new VC account, searches for it, finds nothing, creates it, then chooses whether to add users now by CSV upload or one at a time, or come back later. Each branch became a screen.
Summarized from the same file
Outcome
FFP Smart launched publicly in June 2026, and FFP Lite (MVP 1) is active for early-access users. The launch was a milestone, not the end: I’m still part of the team. I don’t have adoption or business figures to share, so this section shows the scope, the decisions and the work in progress.
The public launch
Core audiences: from four user types to Investor and Startup
Product decisionComponents on shared tokens, handed to engineering
Design system I built
Changing plans touches money, allowances and data, so I mapped it as ten scenarios before designing screens. Three of them, state by state:
Upgrade, Lite to Budget
Immediate downgrade, with hold to confirm
Payment issue and recovery
The decisions behind them:
An immediate downgrade takes effect at once, so the confirm button must be pressed and held for 1.5 seconds. A quick click won’t confirm, and releasing early leaves the plan unchanged.
Trade-off: one extra second of effort for a deliberate action, in exchange for no accidental downgrades.
Customers can downgrade now, or at the end of the billing cycle and keep their current plan until then. A scheduled downgrade can be cancelled.
Data from the higher plan is retained, with a dismissible warning on the dashboard and on billing. Upgrading again restores access and clears the warning.
Running out of reports and a failed payment share one blocked-report dialog with different reasons and next steps, so the product behaves the same way in both cases.
Checkout, invoices and payment methods stay in Stripe. The design covers the hand-off and the return, so customers always come back to a confirmed state.
After launch, the same system extended to a marketplace connecting startups and investors. My contributions included My Listing, Listing Settings and Collaboration Spaces.
In an AI product, trust is a pattern, not a feature. Define it once and every new surface inherits it.
I also learned to hold fidelity low until the direction is agreed. Polishing the first visual style too early cost a redesign.
© Fact Finders Pro. Screens shown to illustrate my design contribution.
Work / Smartflo
Case study 03, information architecture
I was asked to redesign Smartflo’s dashboard so new customers would stay. I delivered it, then visited customers and redesigned the admin portal around how admins actually work, from research to engineering handoff in 90 days.
Problem
45% of reactive support contacts came in a customer’s first 60 days. The menu mirrored internal modules.
My role
Led the UI/UX redesign across navigation, dashboards, reporting, call records, IVR and user management.
Key decision
Widen the scope from the dashboard to the whole portal, and group it by what admins manage.
Result
Nine sections regrouped into six, a first screen with setup paths, and design to handoff in 90 days.
I owned
Others owned
Problem
of reactive support contacts came in a customer’s first 60 days
CX research, Aug 2022 to Jan 2023. A problem baseline, not a result.workflow areas produced 80% of all support tickets
Same source
Research
Customer visits, mystery shopping, support history and a journey map from the CX team, plus my own audit of every screen and a competitor framework.
Six numbered issues across the header, sidebar and dashboard, before any redesign.
The structural decision
The old menu reflected how the product was built. Settings sat four levels deep, for example Services, then IVR, then Add IVR, then its settings, and Reporting alone held 13 sub-items. I regrouped the portal around the things admins look after.
Key decisions
The dashboard was delivered as asked. But the visits showed the drop-off came from finding things, not from the dashboard. I widened the scope to the whole admin portal within the same 90 days.
Admins look after users, calls, services and reports. The new groups follow those, so a new feature has an obvious home instead of a new menu.
Trade-off: the labels are still mostly names of things, not tasks. A next step would be testing task-based labels.
A welcome screen with setup paths and a setup checklist replaced the blank first screen, and empty states offer an action instead of an empty table.
Making calls is what the product is for, so Click to Call sits in the top bar on every screen and opens a short form without leaving the page.
Drag the handle to compare.
From any screen, Click to Call opens a quick call form over the current page, so the admin never loses their place.
Mobile app
Management asked to see how the companion app could change. I explored it, but my recommendation was not to prioritize it.
What we saw in research didn’t point to the app. Customers weren’t using it to make calls or to take calls from their own customers. They opened it mainly to check records, such as how many calls an agent had answered.
So the exploration kept the app focused on what people actually did with it, and the case for the real investment stayed with the admin portal, where new customers were getting stuck.
The app as it was
Exploration, at management’s request
Process
Structure was tested on paper first. The first visual draft leaned on illustration; I dropped it for a quieter layout that put the setup paths and the checklist first.
Outcome
I worked with Tata’s engineering team and Acefone, the implementation partner, including an office visit to walk through the designs and work through changes together. I relocated to the US before launch metrics came in, so this case study shows the work, not a scoreboard.
“In a span of 90 days demonstrated her grip on her subject by mining right insights to define action plan to co-create design principles and UI/UX framework for the business unit.”
Vaibhav Ambradkar, my manager at Tata Teleservices
Treat the requested screen as a hypothesis. The structure around it is often the real problem.
Next time I’d agree on what to measure before design starts, so the results can speak for themselves.
Work / Amazon
Case study 04, systems and standards at scale
Amazon Advertising helps brands reach shoppers on amazon.com, in the Amazon app, and on Fire TV and Fire tablets. Over seven years I went from producing ads to leading creative direction and the standards that let many designers ship one quality bar. Three chapters: brand stores, campaigns across devices, and standards and training.
Problem
Brands needed stores that felt like their own on a shared no-code system, and many designers needed one quality bar.
My role
Designed stores and their visual framework; created the review framework and the designer onboarding program.
Key decision
Design one column first, and grow the system only by adding, never by breaking live stores.
Result
22 store pages in 2020, onboarding adopted in Bengaluru and Bratislava, and a DSS Global Award.
Chapter one
A Brand Store is a brand’s own multi-page shop inside Amazon, on desktop and in the Amazon app, built from templates and tiles in a no-code builder. It’s where shoppers browse a brand’s whole range in one place, and a brand can launch one without developers. I designed stores and the visual framework brands built them from.
I owned
Others owned
Each home page started on paper as numbered zones, with rules for what a brand could put in each. The same grid could hold bar tools for one brand and yarn for another.
Every layout had to work stacked on a phone before it became a desktop grid. A store that breaks on a shopper’s phone fails the brand.
Trade-off: some desktop richness.
The canvas moved from gray to white, a smaller tile size arrived, then video, all while stores were live. Every change had to keep earlier stores working.
Proof across brands
store pages designed in 2020
Performance recordsSales and conversion data is confidential, and advertisers can edit stores after launch, so live stores may differ from my designs.
Systems thinking
Three changes landed while I was designing on the system. Each changed a variable in the toolkit; none was allowed to break a store already built.
Chapter two
Ads that live inside products people already use: the Fire TV home screen, a Fire tablet’s lock screen, banners beside search results. Each placement has its own size, viewing distance and way of being used, from a remote control to a touchscreen. Seven campaigns, each adapting one idea to very different screens.
Brief
An interactive Fire TV ad to help people find a Windows laptop that suits them, using an Alexa voice skill.
Insight
Comparing laptops means comparing many specs. Voice offered an easier way in, but it had to be introduced without overwhelming a viewer on the sofa.
Solution
A three-step journey: an inline banner introduces the finder and the voice skill, a short video adds context, and a Fire TV landing page explains how to use it. Soft colors and restrained type keep it calm at 10 feet.
Outcome
A connected journey that takes a viewer from a single banner into voice-led product discovery.
Brief
Beiersdorf wanted NIVEA’s existing ads to feel more premium, with each product standing out and the brand still instantly recognizable.
Insight
The old ads grouped every product on a light background, so none of them stood out. NIVEA’s signature blue was the strongest brand asset, and it was underused.
Solution
I gave each product its own architectural arch on NIVEA blue: strong contrast against light packaging, a more curated feel, and bold white type for the offer.
Outcome
The new approach was later extended to more European markets.
Brief
A premium holiday campaign for ghd’s gifting collection, showing several products without losing the brand’s luxury feel.
Insight
Generic holiday decoration would cheapen the brand. The aim was to make each product feel like a gift, while keeping a whole collection easy to browse.
Solution
Deep red, gold accents and lit podiums for a festive but luxurious mood. On the Fire Tablet wake screen, tabs let people switch between products in one placement, and circular framing focuses on one product at a time.
Outcome
One festive visual language, working from a tablet wake screen to Fire TV and display ads.
Brief
A creative pitch for Barilla’s Vero Gusto, a premium pasta sauce made with Italian ingredients and traditional recipes, across Amazon placements in 24 hours.
Insight
Its strongest differences were Italian origin, natural ingredients and a homemade character. A recognizable Italian setting could carry that authenticity better than product shots alone.
Solution
I explored concepts and chose an Italian architectural setting with a restrained palette and product-led composition. A junior designer built the banners from that direction; I reviewed them and selected the finals.
Outcome
A presentation-ready pitch across several placements, delivered within the 24 hours.
Brief
Nine display sizes, a landing page and sample packaging, each carrying the product, the claim and several brand logos.
Insight
Some units were as small as 320 × 50 pixels, and the client wanted every brand asset kept. The challenge was hierarchy, not decoration.
Solution
I first simplified the smallest sizes by dropping secondary logos. When the client asked to keep them, I solved it with type, spacing and layout instead, leading every size with one claim: “Replumps in 1H.”
Outcome
A consistent system across nine sizes, a landing page and packaging, through several rounds of client feedback.
Brief
Creative for two Pixel launches across markets and placements, keeping Google’s identity recognizable while each product had its own message.
Insight
Consistent brand doesn’t mean the same approach every time. Pixel 7a needed clear product visibility and localized copy. Pixel 8 Pro’s pre-launch needed anticipation without revealing the phone.
Solution
For Pixel 7a I worked on Japanese-language ads, inline banners and grid layouts. For Pixel 8 Pro, the team explored a pre-launch direction built on elegance and craft.
Outcome
Contributions to a global brand’s campaigns, balancing brand consistency, localization and production constraints.
Brief
An internal challenge among senior designers to choose contributors for the first phase of Amazon Brand Stores: three formats in 24 hours.
Insight
Three formats isn’t one creative resized. Each placement has its own layout, viewing distance and message.
Solution
One idea, “Z50: insanely small, amazingly bold,” built around the camera’s size, then adapted to each format through discovery, ideation, visual design and a final quality check.
Outcome
Selected from the Bengaluru senior designer group for the first phase of Amazon Brand Stores.
Chapter three
How one quality bar held across many designers and many brands: master templates that produce every standard size from one file, a review framework everyone shared, and an onboarding program adopted in two design hubs.
Master templates
One master file produced every standard display size. Designers changed the content once and every format followed, with safe zones, type hierarchy and brand rules already built in. Time went into the idea instead of rebuilding the same ad six times by hand, and fewer errors crept in between sizes.
Template examples shown here are recreated for portfolio use; original working files are NDA-protected.
The system in practice: executions across brands and formats.
Criteria that adapted to each execution type, covering hierarchy, layout, typography, safe zones and CTA destinations. Designers used it to self-check before submitting, and reviewers used it to give consistent, less subjective feedback. It’s what a design system’s guidelines do: a shared definition of “right” that holds without the author in the room.
New hires used to get a month of training across every execution type before using any of it. I proposed and built a new sequence and refined it with new-hire feedback over about two years. It was adopted for design new hires in Bengaluru and Bratislava.
Global Award, Best Internal Reviewer (Bengaluru)
Award recorddesign hubs adopted the onboarding program I built
Bengaluru and Bratislavajunior designers whose work I led
Without a formal reporting lineA good system gives everyone the same structure and still lets each result feel like its own.
Templates, review criteria and onboarding are the same kind of object as a component library: structure that makes every later decision easier. Adoption, not existence, is what makes them work.
About
I trained in painting, printmaking, composition and photography, with a master’s thesis on Kangra paintings. That foundation still shapes how I see hierarchy, balance and detail.
My career grew from web and mobile design into startup product work, then seven years at Amazon in advertising and creative leadership, enterprise UX at Tata Teleservices, and early-stage product design at Fact Finders Pro. I joined Fact Finders Pro because its focus on credible, verifiable information genuinely interested me.
I was the first woman among the daughters in my extended family to pursue higher education and a professional career. Next, I want sustained ownership of a meaningful product area: researching it, designing it, seeing it built and improving it.
Where it began
At art school I kept switching mediums: painting, lithography, etching, and photographs printed by sunlight. Etching taught me the lesson I still design by. A print is built in layers, and a flaw in the plate shows up in every print you pull.
“Plan the plate before you pull the print.”
The rule I carried from the print studio into product design
At Amazon I trained new hires from 2018 and later owned onboarding planning for design new hires. I especially enjoy mentoring designers moving into UX from visual design or fine arts. I know that path well.
“I appreciate her boldness and confidence that helped her grow in her career path of artist (fine arts) to Creative Lead.”
“Her outstanding communication skills, coupled with her open-mindedness and receptiveness to feedback, make her an invaluable team member.”
Outside work
I still paint in watercolor, which teaches planning ahead because you can’t undo a wash. I cook home-style Indian food, and I hike whenever I can get to the mountains, all the way up to Annapurna Base Camp.
Before the work in these case studies, I designed interfaces for 11 iOS and Android client apps and 3 web products, and at Tymtix designed the SensAiry and Bird App mobile apps.