Campaigns and brand work for 150+ brands, including Netflix, Prime Video, Microsoft and LEGO.
Nov 2021 – Jun 2023
Executive – UI/UX, Centre of Excellence
Viatris Inc. (Mylan + Pfizer Upjohn)
Three awards in 2022: Excellence, Brilliance and Ascent.
Dec 2023 – May 2024
Product Designer
Onsurity Technologies Pvt. Ltd.
Product and marketing design for a health insurance platform.
May 2024 – Present
Consultant – UI/UX Design, GLCS
Eli Lilly and Company
Enterprise learning and accessibility. Kingfisher Award 2025 for LUMA.
My work
Two sides of how I create
Structured problem-solving for real products, and colour and brushwork just for the joy of it. Step into either room.
Kind words
People said nice things
Managers, teammates and friends on what it's like to build things together.
01 / 07
“
Atika, throughout the past year, your dedication, creativity, and unwavering commitment to excellence have brought immense value to our work. From tackling challenging projects to consistently delivering high-quality results, your contributions have been nothing short of remarkable.
It has been a pleasure to see your growth and achievements over the past year. Your ideation and solutions have helped us significantly impact our team's success.
GSGarima SrivastavaCurrent Manager“
Atika was one of the most energetic and driven individuals on our team. Her exceptional skills, creativity, and positive mindset shine through in everything she does.
Her collaborative nature, excellent communication skills, and approachable personality make her a true asset. She brings enthusiasm, innovation, and a can-do attitude to every project.
AKAnant KulkarniUI/UX Designer · Mentor“
Having worked with Atika in two different companies, I can attest to her exceptional skills, creativity, and professionalism.
Atika blends aesthetic appeal with functional design, creating stunning and intuitive user interfaces. She consistently delivers high-quality work on time and within budget.
JLJoyal LoboManager, Eli Lilly“
Atika brings such great energy to every project. It's genuinely uplifting to work with her. She's always excited to take on something new and is constantly looking for better, smarter ways to get things done.
She's the kind of person who jumps at opportunities and always follows through, delivering solid work right on time. Working with her is not just productive, it's also a lot of fun.
SSSaloni SharmaBrand Designer, Atlan“
I've known Atika since childhood, and we also studied together, so I've had the rare opportunity to witness her creativity evolve over many years.
She brings together deep design expertise with empathy and precision, whether it's UI/UX, motion graphics, or accessibility-first design. Any team would be lucky to have her on board.
BRBianca RiatProduct, Healthtech“
Atika has always been one to learn, apply and grow. Her attitude towards life and professional space demonstrated strong collaboration, technical know-how and positivity since her early years.
6+ years designing enterprise software and accessible products. AI enthusiast and lifelong learner: I'm always researching, testing new tools and picking up the next technical skill.
UX & productAI enthusiastAlways learning
Hey, that's me! Click for About me
Scroll to step insideSwipe to step inside
atika ajmera
Product & UI/UX Designer● Bengaluru, India
Open to work · Bengaluru
0%
🏁
About me
Open to work
Hi again. I'm Atika.
Product Designer with nearly 6 years of experience in UX/UI design, design systems, and user-centric digital products. Seeking opportunities in Visual Design, Product Design, and UI/UX Strategy & Experience Design to create scalable, accessible, and impactful user experiences that drive business growth and customer satisfaction.
“The best design is invisible. It simply works, for everyone.”
How I work
Design systemsEnterprise systems used across 50+ healthcare brands.
Accessibility firstWCAG built into decisions from the very start.
Research to shipFrom research and prototypes to polished interfaces.
AI-assistedAI to speed up thinking, with human judgment where it matters.
Visual Studio · LDR · Lilly backendData never leaves Lilly infrastructure
GLCS Project Hub · Team lead
The team lead’s home screen, with the read-only client tracker in front.
₹1.2 CrSaved every yearTotal annual cost saving from moving off Jira and streamlining operations
₹30–40 LJira licences retiredAnnual Jira spend no longer paid
60–70Live projects, one viewEvery course, its stage and its burn, at any moment
0External vendorsOwned in-house, so changes ship on our schedule
01 · Context
Lilly’s in-house learning studio.
GLCS designs and builds e-learning for L&D partners across Lilly. At any moment, 60–70 courses move through a 14-step production workflow, from effort estimation and content curation, through Alpha, Beta and Gold reviews, to translation and accessibility testing. Each project draws on four hour pools (instructional design, media, QA and project management) and on people shared across many projects at once.
The production lifecycle every project flows through
Effort estimationScope and hours per pool
Intake & curationContent and high-level design
Design & developmentID + media build
AlphaReview, fixes, validation
BetaReview, fixes, validation
GoldFinal hand-over
TranslationLocalised versions
AccessibilityA11y, device and browser tests
02 · The problem
Everyone had the data. No one had the picture.
Projects were tracked in Jira and supplemented by spreadsheets and a first-generation dashboard. The information existed, but it was scattered, expensive to keep, and shown the same way to every person regardless of what they needed to decide.
P1
Which project is at which stage?
Answering it meant opening tickets one by one or asking around. With 60–70 projects live, nobody held the full picture.
P2
How much budget have we burned?
Hours used against the ID, Media, QA and PM estimates lived in spreadsheets, so overruns surfaced after they happened.
P3
Who is over-occupied, who is free?
Work was assigned from memory. Some people were stretched past capacity while others sat under-used, and revenue suffered.
P4
One view for every role
Everyone saw everything, and nobody got what they needed. Designers waded through portfolio data; leads lacked roll-ups.
P5
No view of what’s coming
Expected projects from L&D partners weren’t visible, so resourcing was reactive instead of planned.
P6
Clients asked by email
Partners had no way to check which phase their course was in, so status requests became a steady stream of messages.
Plus cost and control. ₹30–40 lakh a year in Jira licences, confidential project data on a vendor platform, and every workflow change waiting on vendor configuration.
03 · The old design
Version 1 tracked everything and answered very little.
The first-generation GLCS Project Overview Dashboard sat next to Jira and the spreadsheets. It had six screens for project managers (team bandwidth, utilization, active and closed projects, the pipeline and an hours report) and four for team members, swapped with a prototype toggle. The data was there; the answers were spread across screens.
Version 1 · Active projects
1234
A toggle, not a role. Project manager and team member views were swapped by a prototype toggle, not decided by who signed in.
Six separate screens. Status, people and hours lived on different pages, so one question meant several visits.
Manager first, portfolio never. The page opened on project manager cards; there was no summary of all projects.
Details behind a click. Hours by pool and the gate dates sat inside cards that opened one at a time.
Team utilization Hours by phase as a static bar chart, with no thresholds for over- or under-use.
Upcoming pipeline Expected projects as a table of numbers per quarter.
My timesheet The team member view: a weekly grid of hours per project and workflow step.
04 · Research
We studied the work before we drew a screen.
Six methods, one question: what does each role need to decide, and when?
Interviews
Stakeholder interviews
18 sessions across team leads, PMs, instructional and media designers, developers, QA and L&D partners.
Mapping
Workflow audit
Mapped the 14-step production process, every hand-off and the questions asked at each gate.
Audit
Jira usage audit
Compared the features we paid for with the ones the team actually used, and the workarounds around the gaps.
Evaluation
Heuristic review of v1
Assessed the first dashboard against usability heuristics and WCAG 2.2 AA.
Benchmark
Competitive scan
Jira, Smartsheet, monday.com and MS Project: how each handles capacity, budgets and client access.
Card sort
Module card sort
12 participants grouped dashboard modules by what they need daily, weekly and never.
“Status” meant something different to every role.
Status is role-shapedLead: portfolio health. PM: gates and budget. Designer: what’s due. Client: which phase.
Utilization is the revenue leverIdle hours and overbooked people both cost money; the lead must see both before assigning.
Our workflow is specificFourteen fixed steps with Alpha, Beta and Gold gates: a generic issue model forced workarounds.
Planning needs a future viewExpected projects per quarter, per requestor, make resourcing proactive.
Confidential by defaultProject and people data should stay inside Lilly infrastructure.
Who we designed for Four internal roles and one external.
TL
Team lead
Owns delivery and capacity
Goal
Keep 60–70 projects moving without burning people out.
Pain
Pieces the picture together from Jira, sheets and stand-ups.
Alongside the interviews, we reviewed the first dashboard against usability heuristics and WCAG 2.2 AA, and looked at how Jira, Smartsheet, monday.com and MS Project handle capacity, budgets and client access. Every finding had to turn into a design decision.
Version 1 audit: what we found, why it mattered, and the decision it led to
What we found in v1
Why it mattered
Design decision
One layout for every role, swapped by a toggle
Each role asks a different first question
A home screen per role, chosen at sign-in
Status, people and hours on separate screens
Leads pieced the picture together from several places
Portfolio health, capacity and budget on one screen
Project details in cards that open one at a time
Projects you can’t see together, you can’t compare
A stage board that shows every project at once
Utilization as hours by phase, no thresholds
It never showed who is over-occupied and who is free
A per-person heatmap that flags both
Pipeline as a table of numbers
Demand was hard to read by quarter and requestor
A forecast of expected projects per quarter, per requestor
No view for clients
Partners emailed for every status update
A read-only client tracker
Patterns we chose Small, repeatable pieces every role home is built from.
Answer first
A headline that answers
Each home opens with one sentence that answers the role’s main question, such as “3 people are over-occupied and 5 projects are past 80% of budget.”
One action
One primary button per role
Show at-risk projects, add a project, log today’s hours, show blockers, message the project team.
Signals
Counts with direction
Summary cards show the number and its change since last week, so movement is visible, not just totals.
Pipeline
Every project in one bar
A segmented bar shows how many projects sit in each stage; picking a stage filters the list.
Clients
A stepper, not a spreadsheet
Partners see their course’s stages, the milestones and when their review is needed, and nothing internal.
Themes
Light and dark
Every screen works in both themes, from the same set of tokens.
06 · Build vs buy
Jira tracks issues. GLCS needed to run a production studio.
Jira is excellent for software teams. Our work is different: fixed review gates, hour pools per discipline, shared people and external clients. We compared what each option gives us on the questions that matter to GLCS.
Comparison of Jira Cloud and the in-house GLCS Project Hub
Criteria
Jira (Cloud)
GLCS Project Hub
Annual cost
₹30–40 lakh in licences; billed per seat for every licensed user
No licence fees, part of a ₹1.2 Cr annual saving
Workflow fit
Generic issues and boards; GLCS steps rebuilt with custom fields and statuses
The 14 GLCS steps and Alpha / Beta / Gold gates are native
Hours budget by discipline
Time tracking and budgets via paid Marketplace add-ons
ID, Media, QA and PM pools with live burn and 80% early warnings
Resource utilization
Capacity planning sits in higher tiers or add-ons
Over-occupied vs available, per person, per week
Project pipeline
Roadmaps of known work, not incoming demand
Expected projects per quarter, per L&D requestor
Role-based views
Dashboards configured by hand, per user
The view changes automatically when the role changes
Client engagement
Needs licensed seats or a separate service desk
Read-only phase tracker for L&D partners
Data and confidentiality
Stored on the vendor’s cloud
Stored in LDR, connected to Lilly’s backend
Dependency
Pricing and roadmap set by the vendor; list prices rose again in 2026
Owned in-house; no vendor in the loop
Jira characteristics from Atlassian’s public 2026 Cloud plans and pricing. Cost figures are GLCS’s own.
₹1.2 Cr saved every year, and the data never leaves Lilly.
Jira licences retired
₹30–40 L
Operational streamliningUtilization gains, fewer coordination hours, no add-ons or vendor admin
₹80–90 L
Total annual saving
₹1.2 Cr
07 · Approach & strategy
Role first, screens second.
Instead of redesigning version 1 screen by screen, we started from the roles and the decisions each one makes, then rebuilt the screens around those decisions.
How we worked
DiscoverStakeholder interviews, a workflow audit of the 14 steps and a Jira usage audit.
DefineKey insights, five role profiles and the build vs buy case.
ConceptOne source, five lenses, and the anatomy of a role home screen.
Design & prototypeRole-based screens in a clickable prototype, in light and dark themes.
Build & hand-offHand-off for the Visual Studio build, with a role-scoped API on LDR.
What we kept, changed, added and removed Version 1 had the right data. The strategy was to keep it and change how it reaches each role.
Keep
Hour pools for ID, Media, QA and PM
Alpha, Beta, Gold and hand-over dates
Weekly timesheets
Bandwidth per person
Change
Hours by phase → a per-person utilization heatmap
Pipeline table → a demand forecast
Collapsed cards → a stage board
One layout → five role homes
Add
A read-only client tracker
Budget warnings at 80%
An at-risk filter
Change since last week
Remove
The prototype view toggle
Jira licences
Status-update emails
Four rules we held every screen to
01
One source, many lenses
A single data model; the role decides the lens. No duplicate trackers, no copy-paste status.
02
Status in five seconds
The first screen answers the role’s main question before any click.
03
Capacity before commitment
Show who is free before a project is assigned, not after someone is overbooked.
04
Confidential by design
Data scope follows the role. Clients see their courses; nobody sees everything.
08 · Concept
One source, five lenses.
One project model sits underneath everything. Each role looks at it through a lens that answers its own first question, and the lens is decided at sign-in.
One project modelProjects, hours, people and pipeline · stored in LDR
TLTeam leadIs the portfolio healthy?Pipeline, utilization, budget burn, forecast
PMProject managerWill my projects hit their gates?Milestones, hour pools, team
DSDesignerWhat do I work on now?Assignments, bandwidth, timesheet
DVDeveloperWhat blocks the next gate?Build queue, accessibility and device defects
CLClientWhere is my course?Phase, milestones, review windows
Anatomy of a role home The team lead’s home, the same structure every role follows.
GLCS Project Hub · Team lead
12345
The lens. A tab in the prototype; in the product, the role comes from sign-in.
The answer. The role’s main question, answered in one sentence.
The next step. One primary action, the thing this role most often does next.
The signals. Five cards with counts and their change since last week.
The whole picture. Every live project across the stages, filterable by stage.
09 · Information architecture
Role changes the view. Nobody gets everything.
We mapped every module to the roles that need it. The result is a permission model and a navigation model in one.
Module access by role
Module
Team lead
PM
Designer
Developer
Client
All every project Own only the person’s projects or tasks Read view without editing — no access
10 · How it’s built
Inside Lilly, end to end.
Confidential project and people data stays on Lilly infrastructure. The role from sign-in scopes every API response, so a view can’t show what a role isn’t allowed to see.
Step 1 · IdentityLilly sign-inSingle sign-on returns the person and their role
Step 2 · Front endGLCS Project HubBuilt in Visual Studio; role-based screens
Step 3 · ServiceRole-scoped APIEvery response filtered by role and ownership
Step 4 · StorageLDRProjects, hours and pipeline stored on Lilly’s side
Step 5 · SystemsLilly backend serverConnected for people and organisation data
11 · Design
From one dashboard for everyone to the right dashboard for each.
Before · v1 dashboard
One layout for PMs and team members, switched by a prototype toggle
Project details hidden behind collapsed cards, one open at a time
Utilization as a static bar chart, no thresholds for over- or under-use
Upcoming pipeline as a table of numbers per quarter
No view for clients; status shared by email
After · GLCS Project Hub
A home screen tailored to each of five roles
A stage board that shows where every project is at a glance
A heatmap that flags over-occupied and available people, week by week
Budget burn per hour pool, with warnings at 80%
A read-only client tracker with phases and review windows
Version 1 and the new hub, side by side Drag the handle, or use the arrow keys, to compare.
Version 1New hub
The redesigned screens One product, five home screens. Pick a role.
GLCS Project Hub · Team lead
The whole team lead view
Everything a lead needs to decide before a stand-up, on one scroll: portfolio health, the stage pipeline, team utilization, budget burn and the demand forecast.
Scroll inside
12 · Outcomes
What changed.
₹1.2 CrSaved annually
₹30–40 LJira licence spend retired
60–70Projects tracked in one place
100%Data held in Lilly infrastructure
40%Less time spent on status reporting
12 hCoordination time saved per week
The most valuable design decision was what each role does not see.
Design the permission model with the UI
The role matrix became both navigation and access control: one artefact, two jobs.
Make capacity visible early
Utilization only changes behaviour when it’s on screen before the assignment, not in a monthly report.
Build vs buy is a product decision
Owning the tool let us shape it to our 14-step workflow instead of bending the workflow to the tool.
GLCS Project Hub · live prototype
Switch roles and themes inside the prototype. Data is illustrative.
Case study 02
Persona and mascot design for Global Learning Content Services at Eli Lilly
Meet Luma. Giving Lilly’s learning a face, a tone and a voice.
How a two-week avatar challenge became the identity of Lilly’s message-based learning experience, and a teammate learners recognise.
My role
Media designer · co-creatorVisual identity, brand and emotional direction, production assets
Partner
Diksha RoyInstructional designer · persona and learning design
Timeline
19 June – 4 July 2025Challenge to Gold in about two weeks; mock-ups 16 July
Tools
Adobe Firefly (GenAI) · PowerPointDelivered as PNGs, GIFs, persona deck and device mock-ups
Luma on mobile: the persona carries every nudge, reminder and celebration.
1 of 5Concepts selectedLuma (Option 2) was the stakeholder favourite at Alpha review
15 daysChallenge to GoldFrom brief on 19 June to final assets on 4 July 2025
4.0 ★Teams app ratingFrom 51 ratings on the Luma app listing
50+Installs in a monthShown on the Microsoft Teams app listing
01 · The brief
A new learning tool needed someone to speak for it.
In June 2025, GLCS leadership set the Media Design team a challenge: create a persona for Arist, a new message-based learning tool that would reach employees directly in Microsoft Teams. Each media designer partnered with an instructional designer.
Follow Lilly brand guidelines
Meet accessibility guidance
Media + instructional design pairs
GenAI and Adobe Firefly encouraged
The brief asked for an avatar. We set out to design the emotional layer of the product.
02 · The problem
Message-based learning can feel mechanical.
Nudges and lessons arrive in the middle of someone’s working day, in the same app as their meetings and chats. Without a recognisable voice, they read like system notifications: easy to ignore, easy to resent.
P1
Learning felt mechanical
People learn better when the experience feels human, not automated.
P2
No consistent face
Without a persona, every course, nudge and reminder spoke with a different voice.
P3
High-stakes moments
From frontline operators to office teams, people respond better to empathetic guidance under pressure.
P4
A tiny canvas
The persona had to work as a 32 px avatar in a busy Teams chat list.
How can the personality of the learning experience itself represent Lilly?
03 · Research
Designing for a moment, not a screen.
We started with the learner’s context: someone opening Teams on a difficult day. What would make a learning message feel like a small moment of calm rather than one more demand?
Brief
Brand and accessibility guidance
Lilly brand guidelines and accessibility requirements framed the visual and language choices from day one.
Learners
Context of work
Frontline operators, lab and office teams; onboarding through advanced training; busy shifts and limited attention.
Learning science
How nudges work
Cognitive-load reduction, spaced learning, behavioural nudges and learner engagement.
Platform
Teams constraints
A circular avatar, recognisable at about 32 × 32 px, with master artwork at 500 × 500 px or above.
Values
Team Lilly values
Help People, Make Team Lilly Better, Deliver Results: how could a character carry them?
Stakeholders
Alpha review feedback
Five concepts reviewed; stakeholders shared the characteristics they wanted before we closed the design.
The persona had to earn attention, not demand it.
Familiarity builds trustA consistent face across every touchpoint creates recognition and connection.
Tone matters most under pressureIn high-stakes work, an empathetic voice makes critical training feel approachable.
Small size, big recognitionThe character must read instantly in a chat list: silhouette over detail.
Values, not decorationTo feel like Lilly, the persona needed to embody its values rather than its logo.
Where Luma helps across the learning journey Principles applied: cognitive-load reduction · spaced learning · behavioural nudges · learner engagement
A name that means light. A personality that means care.
“Luma” comes from lumen, a unit of light. The concept: a guide who brings clarity, not just answers; reassurance, not just instructions.
Luma
Illumination
Clarity
Guidance
Learning
Five traits, deliberately balanced The same Luma on day one of onboarding and in a tenth compliance refresher. Pick a trait to hear how it sounds.
A culture carrier Mapped to Team Lilly values.
Value
Make Team Lilly Better
Inclusive, clear and proactive.
Value
Help People
Encourages learning with heart and humanity.
Value
Deliver Results
Keeps messages focused, relevant and action-ready.
05 · Visual design
Friendly, professional, and readable at 32 pixels.
I set the visual direction to make Luma feel like part of Lilly rather than a generic digital assistant: warm enough to smile at, professional enough for compliance training.
Circular avatar
Built for the Teams avatar frame from the start.
Semi-flat, clean lines
Minimal detail that survives scaling down.
Lilly palette
A signature red hoodie with the Lilly script.
Calm, friendly expression
Approachable without being overly casual.
High contrast
Dark frames, warm hair and red read clearly at any size.
32 px to 500 px+
Recognisable as a chat avatar; crisp as master artwork.
One character, many moments The expression set, animated for messages, videos and announcements.
HelloWelcome and onboardingYesGuide: lessons, tips and walkthroughsCareValues, wellbeing, thank-you messagesBest of luckEncourage: milestones and “you’ve got this”Heart signCelebrations and recognitionExcitedAnnouncements and short videos
Try it: does she still read at chat size? The silhouette (wavy hair, bold glasses, red hoodie) was designed to stay recognisable at the 32 px size of a Teams chat avatar.
Luma9:02 AM
Just a friendly reminder: your next module is ready whenever you are.
AI-assisted workflow Created with Adobe Firefly, directed by design.
GenerateExplore character directions in Firefly against the brief.
CurateSelect against brand, inclusion and small-size criteria.
RefineStrengthen the Lilly identity and expression range.
ProduceExport PNGs, GIFs, the persona deck and phone, laptop and tablet mock-ups.
06 · Voice and conversation
What makes her a persona, not just an avatar.
With Diksha, we defined how Luma speaks: the quiet nudge before a busy shift, the encouraging note after a long day.
Try it: Luma in Microsoft Teams Pick a moment in the learning journey.
LumaLearning · Arist app
07 · Accessibility and inclusion
Built for everyone, from the brief onward.
Accessibility wasn’t a final check. It was written into the challenge and the persona definition.
Gender-neutral tone
Warmth without stereotype in visuals and language.
High contrast
Clear figure-ground for visibility at every size.
Globally resonant
Designed to work across roles and geographies.
Friendly, not casual
Approachable while staying professional.
Inclusive language
Clear, respectful and culturally aware.
08 · Where Luma shows up
Wherever learning happens.
Teams learning nudges
Message-based microlearning in the flow of work.
Short learning videos
As voice and avatar.
SOP updates
Reminders when procedures change.
ILT and CBT touchpoints
Bridging classroom and e-learning.
Onboarding journeys
A familiar face from day one.
Delivered through Teams and SMS, with automated reminders, the learner’s preferred language and, where applicable, course credit in SuccessFactors.
Across devices Phone, laptop and tablet mock-ups made to support the rollout.
09 · Process
From challenge to capability.
19 June 2025The challengeBrief issued: a persona for Arist that follows Lilly brand and accessibility guidance, with GenAI encouraged.
19 JuneConcept with Diksha RoyOn very short notice, we developed Luma as a complete persona: character, personality, voice, values and example messages.
23–24 JuneAlpha reviewFive avatar concepts presented. Option 2, Luma, was selected as the favourite.
Late JuneRefinementStakeholders shared the characteristics they wanted before we closed the character design.
4 JulyGold: final submissionPNGs, GIFs and the persona deck delivered and approved as good to go.
16 JulyDevice mock-upsPhone, laptop and tablet mock-ups to support rollout across contexts.
LaunchIntroducing LumaA company announcement named Lilly’s Arist experience “Luma” and credited Atika Ajmera and Diksha Roy for designing her and bringing her to life.
PilotsCommercial and Leadership DevelopmentA CLS/Marketing experience and a Leadership Development experience reinforcing Elevating Performance principles.
2026Lilly’s AI-powered learning partnerLuma is presented enterprise-wide, with GLCS intake guidelines, a content curation template and objectives to expand and standardise its use.
10 · Impact
From avatar to enterprise identity.
Luma became two connected things: the name of Lilly’s branded learning platform, and the persona that gives it a recognisable voice.
Creative challenge
Persona
Selected design
Production assets
Pilots
Enterprise capability
Standardised GLCS service
Recognition
Credited company-wide
Named as the designers who brought Luma to life in the launch announcement.
Adoption
4.0★ · 50+ installs a month
From 51 ratings on the Microsoft Teams app listing.
Culture
Brought to life in person
Luma appeared at the Global L&D Conference 2025; colleagues even dressed as her.
Scale
A GLCS service
Intake guidelines, a content template and 2026 objectives to expand utilisation.
Luma, in real life
Global L&D Conference 2025#LumaInAction on Viva EngageMicrosoft Teams app listingLuma, brought to life
11 · After launch
A supportive persona has to behave supportively.
Real learner feedback kept shaping the experience after launch: the most valuable lesson of the project.
8Short Teams lessons in the Leadership Development rollout
<5 minPer lesson, generally ≤2 messages per week
≤3Luma courses per learner at once: a governance guardrail
What learners said
Reminders came across as too enthusiastic and persistent.
What changed
Reminder behaviour was adjusted to be less intrusive and more helpful.
Success isn’t sending more learning. It’s sending the right learning at the right moment.
12 · Reflection
What this project shows.
Visual design
Character creation and brand identity
UX thinking
Designing for stress and limited attention
Conversational design
Voice, tone and messaging patterns
Accessibility
Contrast, small-size legibility, inclusive language
AI-assisted design
Adobe Firefly in a directed creative process
Instructional design partnership
Co-created with Diksha Roy
Stakeholder iteration
Concept, Alpha selection, refinement, Gold
Enterprise scale
An identity adopted across a global learning platform
I didn’t make a mascot. I co-designed the identity and experience of Lilly’s digital learning companion.
Next, I’d explore
An expression and motion library mapped to message types
Learner sentiment tracking tied to Luma’s tone
Voice guidelines for localisation across languages
Case study 03
Self-initiated project on caregiver medication management
Tend, for the person who keeps track of someone else’s medicine.
A mobile app for adults who manage doses and doctor visits for a parent, as well as their own. Designed and built as a working prototype.
My role
Research, UX, UI, prototypeSolo project, end to end
Type
Self-initiatedNo client or live product; assumptions are labelled as such
Domain
Healthcare · caregivingMedication and appointment management
Output
Working prototypeLight and dark themes, seven testable flows
Today in both themes, and the full-screen reminder.
6Screens, down from 13The same scope as the baseline concept I audited, plus sheets
1 tapTo move a late dayTake a dose late and the remaining doses shift by the same delay
0Accounts requiredEverything works on the phone. An account only adds sync and booking
PlannedUsability testingFive caregiver sessions and a survey are the next step
01 · Why this project
Medication apps are built for one patient. Care is often done by someone else.
I have spent most of my career designing healthcare and insurance products. Medication apps tend to assume a single person managing their own treatment. I wanted to design for the person they treat as an edge case: the adult son or daughter who keeps a parent’s medicines and appointments on track, usually from a distance and alongside a full working day.
P1
Alarms say when, never whether
A phone alarm cannot tell you if the dose was actually taken, by you or by your parent.
P2
A late dose means arithmetic
Ninety minutes late means working out where the next three doses move to, then editing three alarms.
P3
Family set-ups meet a paywall
Established apps put extra profiles, refill alerts or history behind a subscription.
P4
Health data behind a sign-up
People are wary of handing a parent’s health details to an app that demands an account first.
How might one person keep two people’s medication accurate, without accounts, arithmetic or a paywall?
02 · Approach
What I did, and what is still to be validated.
Without a client or a live product, I was strict about separating evidence from assumption. Everything below is labelled as one or the other.
Done
Desk research on medication adherence
Secondary reading of a small published survey
Review of three established apps
Heuristic audit of an existing concept app
Proto-personas, journey, storyboard and flow
Design system and working prototype
Still to run
My own survey with people who manage a family member’s medication
Five moderated usability sessions on the prototype
One iteration based on what breaks
03 · Research
Forgetting is common, and reminders help, with limits.
WHO, 200350%
Average adherence to long-term therapy in developed countries
The World Health Organization’s report puts adherence among patients with chronic disease at only half, and lower in developing countries. Source
BMJ Open, 2020≈2×
Without an app1.0
With an app2.12
App users were about twice as likely to report taking medication as prescribed
Odds ratio from a meta-analysis of nine randomised trials. The authors urge caution: mostly self-report, small samples, 16 weeks of follow-up at most. Source
Adherence to blood-pressure medication, by country
Share of patients adhering to treatment
United States51%
China43%
Gambia27%
Bars drawn on a 0 to 100% scale. Source: WHO, 2003.
Early signals from a small published survey Secondary data. I did not collect it: a 12-person survey from a published student case study, informal and not specific to caregivers. It will be replaced by my own findings.
When you cannot take a dose on time, how much do you struggle to remember it later?
12 responses · 1 low, 5 high
13
20
31
44
54
For Tend: eight of twelve chose 4 or 5. This is the strongest signal in the data and the reason snooze and dose shifting are core features.
How comfortable are you entering personal health information into an app?
12 responses · 1 low, 5 high
15
21
34
41
51
For Tend: five of twelve chose the lowest score. The first button on the first screen starts the app with no account.
Which features would be most helpful?
12 responses
Medication reminders11
Calendar10
Appointment reminders10
Refill alerts8
Booking appointments8
Re-reminders when missed4
Export and share4
Wearables4
For Tend: the top four became the four things Tend does without an account. Booking is the reason to add one.
Would you pay for such an app?
12 responses
No11
Yes1
For Tend: core tracking cannot sit behind a paywall, which is where competitors put family profiles and refill alerts.
Survey data: published case study, n = 12 (source link to be added). Charts redrawn; interpretation is mine.
Competitive review Each established app solves part of the problem.
App
Does well
Where it falls short for caregivers
Medisafe
Mature reminders, caregiver sharing through Medfriend, dependent profiles
Unlimited medications and unlimited Medfriend and dependent profiles sit behind Premium
MyTherapy
Free, no ads, no login. Alerts repeat until a dose is marked taken or skipped. Monthly report
A pharmacist review notes limited dosing detail for manual and as-needed medication
Dosecast
No registration. Flexible schedules, including intervals counted from the last dose
Profiles for several people or pets, dose history, refill alerts and sync are Pro features
Sources: Medisafe help centre, Healthify NZ clinical review of MyTherapy (Nov 2024), Dosecast App Store listing. Checked November 2026.
The competitor is the alarm app. Tend has to win on what alarms cannot do.
Record, not just remindKnowing whether a dose was taken matters as much as being told when.
Late is the normal caseEight of twelve survey respondents struggle to remember a dose they could not take on time.
Privacy before featuresFive of twelve gave the lowest comfort score for entering health details into an app.
Care for two, freeMore than one person, without an account or a subscription, is where every competitor asks for something.
04 · Baseline audit
Auditing an existing concept before designing my own.
I reviewed a published student concept for a medication companion app against Nielsen’s heuristics. It gave me a concrete baseline to improve on. Four findings shaped Tend.
Finding in the baseline
Heuristic
Response in Tend
Home opens on an hourly grid that is mostly empty
Aesthetic and minimalist design
Today lists only what is scheduled, with the next dose as the largest element
A late dose cannot move the doses after it
Flexibility and efficiency
Taking a dose late offers to shift the rest of the day in one tap
Missed and skipped doses sit in a list with no action
Help users recover from errors
Needs attention resolves each one: I took it, or I skipped it
Status shown by small icons and colour only
Visibility of system status
Every status carries a text label as well as colour
Baseline: the same published case study (title, author and link to be added).
05 · People
One carer, one parent, two very different days.
These are assumption-based proto-personas drawn from the desk research, the secondary data and the competitive gap. They are hypotheses, and I will revise them against my own survey and sessions.
M
Meera, 41
Operations lead. Manages her own daily tablet and her father’s three medicines from across the city.
“Did he take the evening one, or did I just assume he did?”
Goals
See both schedules in one glance. Know a refill is coming before the strip runs out. Keep his visits from clashing with her work.
Frustrations
Alarms tell her when, never whether. Short courses mean four alarms to set and later delete. Will not hand a parent’s health details to an app that demands sign-up.
R
Raghav, 68
Retired bank officer, Meera’s father. Routine-led, careful, not confident with new apps.
“Just tell me what to take now.”
Goals
One clear instruction at the right time. Confidence that he has not taken a dose twice. Not to worry his daughter.
Frustrations
Small text and screens with many buttons. Unsure what a notification wants from him. Updates that move things around.
Meera’s day, when one dose runs late
Stage
Doing
Thinking
Pain today
Opportunity
Set up
Adds herself, her father and his eye drops: four times a day for ten days
“I don’t want to make an account for this.”
Sign-up walls, one alarm per dose
Guest by default. One form creates the whole course
Reminder
Phone buzzes in a meeting
“Not now, but don’t let me forget.”
Dismissed alarms vanish
Snooze that comes back, with only three choices
Late dose
Calls her father at 14:10 for the 13:00 drops
“So when is the next one now?”
Mental arithmetic, editing alarms
Offer to move the remaining doses by the delay
Evening check
Opens the app once more
“Did everything happen?”
No record of what was taken
Progress for the day and a list of anything unresolved
Refill
Notices the strip is nearly empty
“I should have ordered on Monday.”
Finds out too late
Count stock down and warn days ahead
Storyboard: the late dose, frame by frame The one scenario existing tools handle worst.
1. Meera is in a meeting when her phone buzzes. It is time for her father's eye drops.2. She cannot call him now, so she snoozes. The reminder will come back.3. At 14:10 she steps out and calls. He takes the dose while they talk.4. She marks it taken. Tend notices the delay and offers to move the remaining doses.5. One tap, and the evening doses move with it. No arithmetic.6. That evening the day shows as complete, for both of them.
06 · User flow
Four tabs, and one way in.
The baseline spread this scope over thirteen screens. Tend uses six, plus sheets. Month and day views are merged, the three editors became sheets, and notifications moved into the place where they can be acted on.
Rounded boxes are screens or sheets, diamonds are decisions. Back navigation is left out to keep the main paths readable.
Four rules I held the design to
Private until you choose otherwise
The primary button on the first screen starts the app with no account. Sync and booking are the reasons to add one later.
Answer “what now?” first
The next dose is the biggest thing on screen. History and planning live one tab away.
Forgive the imperfect day
Late, snoozed, skipped and missed are all normal states with a way forward. Every action can be undone.
Never rely on colour alone
Status is always written out. Main actions are at least 50 px tall, in light and dark.
07 · Sketches and wireframes
Working out the layout before the look.
Quick lo-fi sketches of each screen from the flow. Two ideas did not survive: actions pinned to the bottom of Today, and an add form sharing the Care screen with its list.
Welcome. Guest entry drawn largest.Today, first idea. Plain list with actions pinned at the bottom.Today, second idea. The next dose becomes a card that carries its own actions.Reminder. Time, medicine, person, three choices.Calendar. Month, three totals, then the chosen day.Care, first idea. List and add form on one screen.
Wireframes The next dose became a card that holds Take, Snooze and Skip, and the add form moved into a sheet so the Care list keeps the full screen.
The first visual pass was calm and flat: stone neutrals, a deep green accent and serif numerals. The main action did not stand out. The final direction keeps the same layout and raises contrast and depth. Translucent cards sit over slow ambient colour, the next dose lives on a dark feature card, and lime is reserved for the single action that matters on each screen.
First passFlat cards, green accent, serif numerals.Final, lightGlass cards, dark feature card, lime action.Final, darkIndigo ground with the same lime accent.
Visual research Six patterns I studied in current health, habit and calendar apps. Thumbnails are my own redrawings of each pattern.
Look
Layered, rounded surfaces
Soft cards on a tinted ground are the current norm in health and habit apps. They read as calm and make grouping obvious.
Taken into Tend: Glass cards layered over soft ambient colour, with a single accent.
Navigation
Bottom bar, three to five items
Thumb-reachable and always visible. Floating action buttons that expand into menus hide the main tasks.
Taken into Tend: A floating pill bar where the active tab shows its label. Add buttons sit inside the screen they belong to.
Calendar
Week strip above, month on demand
A seven-day strip gives orientation without the cost of a full grid. Month views work best as a summary.
Taken into Tend: Week strip on Today, month with status dots on its own tab.
List
Time-led rows with one action
The clearest medication lists lead with the time, then the name, then a single control to mark it done.
Taken into Tend: Time in bold numerals, name and person, one tick. Status as a written label.
Welcome
One promise, one primary action
Good first screens say what the app does in a line and offer a single obvious next step.
Taken into Tend: A one-line promise and Start without an account as the main button.
Reminder
Full screen, few choices
Alarm-style screens work because nothing competes with the decision. Three options is the practical ceiling.
Taken into Tend: Time, medicine, person and exactly three buttons.
Core colours and their roles
Primary, light Main buttons, selected day and tabInk #11140F
Primary, dark Selected states and feature card in the dark themeIndigo #5A48E6
Accent The one action that matters: Take, add, confirm. Taken stateLime #D7F04A
Person 2 Second person, appointment markersViolet #5E4FD0
Ground, dark App background in the dark themeMidnight #0C0E2A
Type One geometric family, Outfit, so pills, chips and labels feel like one system.
17:00
Outfit, medium
Times and headings, set large with tight tracking. The time of a dose is the largest thing on any screen.
Take, Snooze
Outfit, regular to semibold
Labels, names and actions. Tabular figures wherever numbers line up.
Icons One outline set at a single stroke weight, so nothing depends on colour.
Capsule
Tablet
Liquid
Inhaler
Person
Pet
Visit
Reminder
Refill
Snooze
Taken
Private
Today
Calendar
Care
10 · High-fidelity UI
The final screens, in light and dark.
Eight screens captured from the working prototype, with real state: a dose running late, a refill coming up and one missed dose waiting to be resolved.
Use the arrows, swipe or tap a side screen. Switch theme at the top right.
11 · Testing
What I will test, and where the results will go.
Planned
Five usability sessions
Participants who manage medication for a family member. Moderated, 30 minutes, thinking aloud, screen recorded. Seven tasks, from starting without an account to taking a dose late and moving the rest of the day. Measures: completion, time on task, wrong turns and what people say unprompted.
Planned
Caregiver survey
Fifteen or more people who manage a family member’s medication. It replaces the secondary data and tests the riskiest assumptions: whether an offer to shift doses feels helpful or unsafe, and whether “no account” matters more than sharing from day one.
Findings and a before-and-after of the screen that changes most will be added here.
12 · Next
Where it goes after testing.
A simpler single-person mode
Larger type and only the Today screen, for someone like Raghav.
Shared access
So Meera sees when her father marks a dose on his own phone.
A doctor-ready summary
Adherence for the last month, ready to show at an appointment.
The reminder was never the hard part. The record and the recovery are.
Tend · live prototype
Tap Start without an account. Example data; nothing leaves your browser.
02 · Work
Visual Communication
Brand, editorial and campaign work. Placeholders for now.
Sketchbook
I'm also an artist.
Paint, ink and pixels. Where I play with colour before it becomes a system. More at @artxtika.
Deque
