Sudarvel
K
V2.0
23.10.2026
Project: OncoSmart
Year: 2025–2026
Type: e-health · home rehabilitation · mobile
Role: Product design + build
00 / THE STORY
Between the hospital and home.
▶ CLICK THE BOOK TO TURN THE PAGE
TL;DR
Result:
– 7 PAR-Q questions gate every exercise plan.
– Every session is checked: pulse and readiness before, pain after.
– Phase II extends one structure to 4 cancer pathways.
India recorded about 15.7 lakh new cancer cases in 2025. Exercise during treatment has strong evidence, yet most patients don’t do it: rehab lives in hospitals that are hours away, and home guidance is a printed sheet in the wrong language.
I designed and built a bilingual home-rehab app that screens for safety before anything starts, shows each exercise on an avatar, checks pain after every session and stops the plan when pain is too high.
It is 7:40 PM on a Tuesday. Meena, 52, is between chemotherapy cycles. Two weeks ago a physiotherapist showed her a set of gentle exercises at the hospital, 4 hours from home. Tonight, in her front room, she can’t remember if it was ten repetitions or twenty.
Chapter 01 · The real problem
Her daughter bought a pulse oximeter. Nobody told her what number means stop.
Her chest feels tight. She doesn’t know if that is normal or a reason to rest. The leaflet is in English. She reads Tamil.
She doesn’t need a step counter. She needs someone to say: “Your pulse is fine. Here’s the movement. Ten times. Tell me how it felt.”
And on a bad day, she needs it to say stop. That is what I built OncoSmart to do.
Meena is an illustrative scenario written from the study team’s notes. The stills are frames from OncoSmart’s own avatar and oximeter-tutorial artwork.
01.1
The problem in numbers
new cancer cases in India in 2025. Breast, oral, lung and colorectal, the four sites OncoSmart’s pathways cover, make up about 41% of them.
of people with cancer don’t meet aerobic exercise guidelines, though exercise eases fatigue, anxiety and depression.
median one-way travel to treatment for cancer patients in a South Indian centre. 92.7% reported financial hardship.
physiotherapists per 10,000 people in India, an estimated 70%+ of them in cities.
of cancer patients report fatigue, the barrier to exercise patients name most often.
of Tamil Nadu’s people have Tamil as their mother tongue. Home exercise sheets elsewhere come in English, Hindi and Marathi.
of Indian cancer patients are 60 or older, median age 67. Design for older hands and eyes is not optional.
of 69 breast-cancer apps connect users to a professional. Engagement scored lowest on app-quality ratings.
02.1
Research finding table
Source
Finding
Literature
Barriers review
Barriers are contextual: access to and availability of the programme. Patients prefer home-based activity.
Guide · 27 Jun 2025
Add PAR-Q after onboarding. Ask cancer type and surgery. Keep heart rate under 220 − age. Add a pain scale.
Guide · 5 Jul 2025
Make it an app, not a prototype. Pause progress during therapy. Progress stage by stage.
Progress brief
“Patient should not feel what’s the point in using” if only streaks show. Show pain going down.
Earlier video study
Experts wanted a pre-session check and post-session feedback. A video alone couldn’t do it.
Exercise guidance
ACSM 2019, COSA
~30 min moderate aerobic 3×/week + resistance 2×/week. Stop for chest pain, dizziness, SpO₂ ≤ 88%.
App quality
MARS
Rate engagement, functionality, aesthetics and information. Used as my pre-validation checklist.
02.2
The gap: what exists today
Oncology apps track symptoms and help people navigate treatment. Physio apps prescribe exercise but know nothing about chemotherapy. The closest matches are research prototypes abroad. In India, rehab is in person, and home exercises are printed pamphlets.
Product
What it is
Exercise plan
Safety screen
Pain-linked stop
Tamil
Kaiku Health
Symptom monitoring for clinics
Not core
Not found
Not found
Not found
Belong · Careology · OWise
Navigation, tracking, community
Not found
Not found
Not found
Not found
Untire
Fatigue self-management
Light activity goal
Not found
Not found
Not found
Physitrack
General physio platform
Yes, not oncology
Not found
Pain tracked, no stop
Not found
Cancer Exercise App
U. Alberta, in trials
Research prototype
Yes, by cancer type
Not found
Fatigue-adjusted
No
Tata Memorial rehab
In-person service + pamphlets
In person
Clinician
Clinician
EN · HI · MR only
OncoSmart
Home rehab app, physio-led
4 pathways × 4 levels
PAR-Q gate
Above 6 stops plan
Full flow
“Not found” means not found in public product material in October 2026, not confirmed absent. I also found no published Indian trial of an app-based exercise programme for cancer patients.
COM-B: barrier to design lever
Driver
Barrier at home
Design lever
Capability
Can’t remember the movement or reps. Doesn’t know what’s safe.
Avatar demo, big rep count, PAR-Q, pulse check
Opportunity
Hours of travel. Few physios. English-only material.
Home delivery, full Tamil flow, chat support
Motivation
Fatigue, fear of harm, no visible progress.
Ready check, pain trend, badges, pause
02.4
The 5 dimensions I identified
“Heart rate should be maintained below that level using pulse oximeter.”
Study guide · 27 Jun 2025
01 The clinic is the bottleneck
The exercises are simple. Reaching the person who teaches them is not. Every flow assumes the patient is alone at home.
02 Safety before motivation
Heart conditions, dizziness and post-surgical pain change what is safe. Encouragement that ignores this is a liability.
03 Streaks are the wrong signal
A streak says you showed up. It doesn’t say you’re getting better. Pain going down does.
04 Treatment breaks the routine
Chemo and radiation weeks are planned breaks. A plan that punishes them loses the patient.
05 Language is access
A Tamil-speaking patient with an English leaflet depends on someone else to exercise safely.
“Progress paused when the user undergoes any therapy.”
Study guide · 5 Jul 2025
“Patient should not feel what’s the point in using — only streaks are showing.”
Progress screen brief
02.5
The research-to-design strategy map
Research evidence
Core insight
Design decision
4.3 h travel; 0.36 physios / 10k
The clinic is the bottleneck
Full programme at home; physio content, not generic fitness
Guide: PAR-Q after onboarding
Not everyone is safe to start
7-question gate with two exits; no plan before it
Guide: HR below 220 − age
Effort needs a ceiling
Wear-your-pulse prompt (< 110 bpm) + oximeter tutorial
Guide: pain scale in every case
Pain is the main safety signal
0–10 check after every session; above 6 stops the plan
Brief: streaks feel pointless
Show recovery, not attendance
Pain + difficulty charts where lower is better
Guide: pause during therapy
Breaks are part of treatment
Pause Progress; no streak resets
88% Tamil mother tongue
Language is access
Full Tamil flow, written alongside English
Steps and hours mean nothing to someone in the middle of chemotherapy.
The study guide’s note was blunt: a patient who sees only streaks will ask what the point is.
The only number that says “this is working” is pain, and how hard the same exercise feels.
I removed hours, steps and the tracker framing. Not deprioritised. Removed.
The same review replaced my age and gender dropdowns with large chips. Dropdowns hide the options and give small targets to older hands.
04.2
Is / is not table
OncoSmart is
OncoSmart is not
A physio-led home programme
A generic fitness tracker
A safety screen before anything starts
A sign-up disclaimer
A guide that shows and counts
A video library
A record of pain and effort
A streak game
A route to a human when it hurts
A symptom diagnoser
Tamil from the first screen
An English app with captions
04.3
5 non-negotiable principles
01 Screen before you start
No exercise content before the PAR-Q is complete. Enforced by route order, not by a reminder.
02 Every session ends with a question
A session isn’t complete until pain is recorded. The pain sheet closes the session.
03 Recovery over activity
No steps, calories or minutes. Every chart shows a falling line, and says that lower is better.
04 Pause is not failure
Treatment weeks freeze progress instead of breaking it. Nothing is lost.
05 A human is one tap away
The chat button sits on every main screen. Discomfort is reported in one tap, in either language.
Motivation permitted in
Motivation forbidden in
Home quote card
PAR-Q questions
Day-completed state
Consult-your-doctor exit
Badges and their copy
Pulse and ready checks
Rest screen
Pain score sheet
Level ring
High-pain stop
Chat when reporting discomfort
05.2
The four system laws
A “yes” is a real outcome. The plan doesn’t open.
Showing up is not the same as getting better.
Treatment weeks freeze the plan. They never reset it.
Above the clinicians’ threshold, the plan stops and a person takes over.
Any push to continue on a safety surface is a breach, however kind the words are.
Object
States
Rule
Patient
onboarding → screened → cleared / referred
No plan without screening
Pathway
breast · head & neck · thorax · abdomen
Set by cancer type; defines exercise set
Level (4)
upcoming · active · paused · done
Visible ahead; progress stage by stage
Day (5 per level)
to do · done
One session per day
Session
ready check → playing ⇄ rest → pain recorded
Not complete until pain is recorded
Exercise
demonstrator × language × reps
One player template
Pain entry
0 · 1–3 · 4–6 · 7–9 · 10
Above 6 stops the plan
Badge (5)
locked · earned
Names a capability, not a count
06.2
Information architecture
Safety gate or check
Stop or refer
Screen
The full structure I mapped before high fidelity. Four tabs plus a floating chat. Safety gates sit on the only paths into exercise and out of a session.
06.3
User flows
Decision
Safe outcome
Stop or refer
07.1
Key design decisions
Decision 01
A Gate, Not a Disclaimer
What I decided
Seven PAR-Q questions on two screens. Any “yes” sends the patient to their doctor. The plan stays closed.
The default direction: A sign-up checkbox, “I confirm I’m fit to exercise.”
Why I rejected it: A checkbox moves the risk onto the patient and trains them to tap through.
Alternatives I considered: A single yes/no question; showing the PAR-Q only after Day 1.
Trade-off I accepted: Two extra screens before any content.
What I gained: Unsafe starts are stopped before they begin, in words the patient can actually answer.
Decision 02
Pain Closes Every Session
What I decided
A 0–10 slider ends each session. Above 6, the plan stops and routes to expert support, full screen.
The default direction: An optional symptom journal.
Why I rejected it: Optional means the people in the most pain skip it. The guide asked for a pain scale in every case.
Alternatives I considered: Weekly pain check; emoji-only mood rating.
Trade-off I accepted: One more required step after effort.
What I gained: The plan knows when to stop. The threshold belongs to the clinicians, and the UI says so.
Decision 03
Two Falling Lines, Not Rising Numbers
What I decided
Progress shows pain (0–10) and exercise difficulty (1–5). Both should fall. Each chart says so.
The default direction: Streaks, minutes, calories.
Why I rejected it: For someone in treatment, “more” isn’t the goal. Feeling less pain is.
Alternatives I considered: Steps and hours (built in low-fi, then removed); a single wellness score.
Trade-off I accepted: Charts where down is good need a label.
What I gained: “Last week I rated it 4. Now it’s 2.” Proof the patient can feel.
Decision 04
Pause, Not Penalty
What I decided
Pause Progress sits under the level ring. Treatment weeks freeze the level and day. Nothing resets.
The default direction: A daily streak that resets on a missed day.
Why I rejected it: Chemo weeks are planned breaks. A reset says “you failed” at the most fragile moment.
Alternatives I considered: Streak freezes as rewards; auto-detecting inactivity.
Trade-off I accepted: Less daily pressure to return.
What I gained: Patients come back to the same place after treatment.
Decision 05
Say What’s Wrong in One Tap
What I decided
Support is a chat with quick replies: Report discomfort → Pain, Tiredness, Breathlessness. In English and Tamil.
The default direction: A help page with FAQs and a phone number.
Why this is a call only a designer close to the patient makes: A breathless patient mid-session won’t read an FAQ or type a symptom in a second language.
Alternatives I considered: Free-text chat only; an AI symptom bot.
Trade-off I accepted: Quick replies narrow what can be said first.
What I gained: Discomfort reaches a person in one tap. The app never interprets symptoms itself.
4 choices that shaped OncoSmart
I wrote down every choice I made, sorted them into big-picture and small-detail, and picked the four that best show how I think. Each one is a short story: the problem, what it was before, what I chose, and what changed.
Tap a row to see the choices in it. Things I chose not to build count too.
I deleted my own progress screen and showed pain instead
Press Next (or the → key) to read it step by step.
08.1 Onboarding
Language
Age
Avatar
One question per screen. Every question has a reason under it, so no one wonders why the app wants to know.
Treatment details
The answer picks the pathway
Cancer type, current treatment and surgery status, all as chips. Cancer type selects one of four exercise pathways. Nothing here is typed.
Choose avatar
A body that looks like yours
The patient follows a demonstrator who looks like them. “Save & Continue” stays disabled until a choice is made, and “You can change it later” lowers the stakes.
08.2 Safety screening
Q 1–4
All no
Decision 01 on screen
Seven questions a patient can answer
Heart condition, chest pain, dizziness, joints, blood-pressure medication, anything else. Plain yes/no pairs, split over two screens so neither feels like a form.
Two exits, one tone. A patient sent to their doctor isn’t made to feel rejected.
08.3 Home and pre-session checks
Home
Pulse
Pulse check
A ceiling before effort
“Make sure your heart rate stays below 110 bpm during the exercise session.” The guide’s rule (keep under 220 − age) becomes one number the patient can watch.
Ready check
“No” is a real answer
The patient confirms they feel ready today. “No” returns them home without guilt. It is the pre-session question the earlier video study couldn’t ask.
08.4 Oximeter tutorial
Step 1 of 4
08.4 UI breakdown · Home screen
What each part of the home screen is doing.
The home screen is the one a patient sees every day, often tired and alone. Every element on it answers one question: what do I do today, and is it working?
The home screen opens on “0 / 5 Days completed” for the current level. One line from research stuck with me: a patient should not feel “what’s the point” if only streaks show. Days in a level show where you are in the plan, and a missed day doesn’t wipe anything out.
Under the progress ring sits one short line, “Stay consistent with your exercises for better recovery”, next to a walking avatar. People between chemotherapy cycles have good days and bad days. The card is a reminder, not a score.
“Today’s Exercise” shows a single card: the day, the avatar who will demonstrate, and one Start button. The plan decides what comes next, so a patient alone at home never has to work out how many reps or which exercise to do.
Only three tabs: Home, Your Growth and Settings. A chat button floats above them, because every flow assumes the patient is alone at home and may need someone to ask.
08.5 The session
Day plan
Ankle pumps
Rest
Day plan
Know the commitment first
“Total Exercise: 19 · Net duration: 32 mins.” A tired patient decides if today is possible before starting, not halfway through.
Player
One number, one instruction
The rep count is the biggest thing on screen. The how-to sits under it in plain words. Pause and Restart are where a thumb reaches. Equipment is household only: a chair, a wall, a bed.
Exercise info
Info sheet
Completed
08.6 Pain check
0 · No pain
10 · Worst
0
No pain
வலி இல்லை
1–3
Mild
லேசான வலி
4–6
Moderate
இடைப்பட்ட வலி
7–9
Severe
கடுமையான வலி
10
Worst pain
மிகக் கடுமையான வலி
Decision 02 on screen
A number, a word and a colour
The dial shows the score, names it in words and turns green, amber or red. “Drag the slider to show your pain score, so we can modify your plan” explains why it matters.
Specified next: above 6, a full-screen “High pain score detected. For your safety, we are temporarily stopping your exercise plan,” with a direct route to expert support.
08.7 Your Growth
Progress · scroll
Workouts
Decision 04 on screen
Pause lives under the ring
Levels completed out of 4, with Pause Progress directly beneath, where a patient starting a treatment week will look for it.
Decision 03 on screen
Lower is better, and it says so
Pain as weekly bars, difficulty as a line. “Lower is better.” “Lower means stronger.” The screen answers the question patients actually ask: is this working?
08.8 Support and Tamil
Chat · EN
Language
Pulse · TA
Decision 05 on screen
Discomfort in one tap
Report Discomfort, then Pain, Tiredness or Breathlessness. Free text is there, but nobody has to type a symptom while feeling unwell.
Tamil
உடற்பயிற்சிக்கான பரிசோதனை
Tamil strings run longer. Titles wrap to two lines and buttons keep their size. Safety copy gets the same care as the main flow.
08.9
Phase II: one structure, four pathways
Breast
Head & Neck
Thorax
Abdomen
Onboarding, screening, the session loop and the pain check are shared. Only the exercise set changes. The breast pathway adds wall climbing, wall slides and a pectoralis stretch for shoulder mobility.
08.11
From Figma to a working app
Design
Figma
Low-fi → hi-fi, EN + TA
Bridge
MCP
Cursor reads frames + tokens
Build
Cursor
AI-assisted, screen by screen
App
React Native
Android + iOS
Data
Supabase
Pseudonymised records
I designed everything from scratch, then connected Figma to Cursor through MCP and built the app in React Native myself. Each exercise is a matrix of videos (male and female, English and Tamil, 5, 10 and 15 reps), so the player is one template with those as inputs.
Next is expert technical validation: layout across devices, readability, media and language fit, progress display and error states on the frontend; pseudonymisation, row-level security, authentication, secrets and backup on the backend.
09.1
What I never built
These are deliberate exclusions, and I’d defend each one. The first was built in low-fi and removed.
Feature
Why it stays out
Steps, calories, active minutes
Activity volume isn’t recovery for someone in treatment
Streak resets for missed days
Treatment weeks are planned breaks, not failures
Leaderboards or patient comparison
Different cancers and treatments make comparison meaningless, and shaming
Auto-progression past pain
The plan never advances beyond the clinicians’ pain threshold
Symptom diagnosis in chat
The app routes discomfort to a person. It never interprets it.
Motivational copy on safety screens
Encouragement there is pressure, however kind the words
10.1
Outcomes
01 / Safety gate
7 PAR-Q questions answered before any plan opens.
03 / Reach
English + Tamil end to end, male + female demonstrators.
02 / Every session
Pulse + readiness before. Pain after. Above 6 stops.
04 / Scale
4 cancer pathways on one shared structure.
These are design outcomes, true by construction. Patient outcomes (activity, quality of life, function) are what the study will measure next. They aren’t measured yet.
The best outcome is Meena on a good evening: she confirms she’s ready, clips on the oximeter and follows an avatar that looks like her through 32 minutes, in Tamil. On a bad evening she rates her pain 7, and the app stops and puts her in touch with a person. Both evenings are the product working.
11.1
What I’d do differently, and what’s still open
01 / The age chips stop at 64
36% of Indian cancer patients are 60+, median age 67. A 65+ band, larger type and a caregiver-assisted setup are my first fixes.
02 / Test with patients
All clinical input came through one team. Next: five patients mid-treatment running Day 1 at home: oximeter, Tamil copy, slider.
03 / Design for drop-off
In one cancer-app study only 54% still used it weekly at week 10, men far less than women. Reminders and physio check-ins need designing.
04 / Ethical pressure
Who sets the pain threshold? “Above 6” is one rule for every pathway. It must be clinician-owned, per pathway, and the consult exit still needs a clearer next step.
Sources
