Mayo Clinic Observership Awardee
Mayo Clinic Observership Awardee
When it hurts at 2 AM,
a pamphlet won't answer.
A digital companion that turns chemotherapy's flood of information into calm, real-time guidance, built with Mayo Clinic.
A digital companion that turns chemotherapy's flood of information into calm, real-time guidance, built with Mayo Clinic.
95%
Task Completion
Task Completion
85%
Engagement Lift
Engagement Lift
30%
Symptom Accuracy
Symptom Accuracy
100%
Red-Flag Detection
100%
Red-Flag Detection
OVERVIEW
OVERVIEW
Chemo patients aren't short on information. They're buried in it, wrong time, wrong format and no one to say if what they're feeling is normal or an emergency. ChemoBuddy is a bedside guide that fixes the timing, not the amount. The right answer, at the moment the symptom shows up, in language a person on chemo can actually process. Built with Mayo Clinic.
Chemo patients aren't short on information. They're buried in it, wrong time, wrong format and no one to say if what they're feeling is normal or an emergency. ChemoBuddy is a bedside guide that fixes the timing, not the amount. The right answer, at the moment the symptom shows up, in language a person on chemo can actually process. Built with Mayo Clinic.
My Role
My Role
UX Designer & Researcher
UX Designer & Researcher
Research, UI Design, Testing, and Iteration.
Research, UI Design, Testing, and Iteration.
Timeline
Timeline
10 Weeks, Fall 2025
10 Weeks, Fall 2025
Client
Client
Team
Team
7 Members
7 Members
Arizona State University
Arizona State University
Highlight
A companion that walks with cancer patients from diagnosis through recovery.
Highlight
A companion that walks with cancer patients from diagnosis through recovery.
PROBLEM
PROBLEM
Core pain points
Treatment is hard enough. The tools meant to help are making it harder.
Treatment is hard enough. The tools meant to help are making it harder.
01
The information exists. The timing doesn't.
Patients get thick binders on day one. They stop reading after five minutes. The real questions come at 2 AM on day four, and the clinic is closed.
01
The information exists. The timing doesn't.
Patients get thick binders on day one. They stop reading after five minutes. The real questions come at 2 AM on day four, and the clinic is closed.
02
Patients can't tell "normal" from "emergency."
Without a reliable way to check, they either wait too long or rush to the ER for something they could have handled at home. Both cause real harm.
02
Patients can't tell "normal" from "emergency."
Without a reliable way to check, they either wait too long or rush to the ER for something they could have handled at home. Both cause real harm.
03
Caregivers are flying
blind.
Family members try to coordinate care with no access to the treatment plan. The patient, already exhausted, ends up explaining their own condition to the people trying hardest to help.
03
Caregivers are flying blind.
Family members try to coordinate care with no access to the treatment plan. The patient, already exhausted, ends up explaining their own condition to the people trying hardest to help.
THE CHALLENGE
THE CHALLENGE
"How might we give chemo patients and their caregivers the right answer at the right moment without increasing cognitive load?"
"How might we give chemo patients and their caregivers the right answer at the right moment without increasing cognitive load?"
RESEARCH
RESEARCH
They had all the content. What they needed was context.
Across 11 peer-reviewed studies, 8 competitive audits, and clinician-guided role-play testing, the same conclusion kept surfacing.
Across 11 peer-reviewed studies, 8 competitive audits, and clinician-guided role-play testing, the same conclusion kept surfacing.
01 - Stakeholder alignment
Three mentors shaped the product. Each gave us a line we kept coming back to.
Three mentors shaped the product. Each gave us a line we kept coming back to.
CLINICAL MENTOR
Dr. Irbaz Riaz
Ask what patients actually need on treatment day, not just how the drugs work
Grounded the case in cabozantinib and nivolumab. Expanded scope beyond medicine to logistics: When do I arrive? How long is the infusion? What foods should I avoid?
CLINICAL MENTOR
Dr. Irbaz Riaz
Ask what patients actually need on treatment day, not just how the drugs work
Grounded the case in cabozantinib and nivolumab. Expanded scope beyond medicine to logistics: When do I arrive? How long is the infusion? What foods should I avoid?
ONCOLOGY MENTOR
Dr. Umar
The chatbot must educate and support, but never diagnose or prescribe
Defined red-flag criteria, fever ≥104°F, chest pain and shortness of breath. Non-negotiable escalation. Every response reinforced the boundary.
ONCOLOGY MENTOR
Dr. Umar
The chatbot must educate and support, but never diagnose or prescribe
Defined red-flag criteria, fever ≥104°F, chest pain and shortness of breath. Non-negotiable escalation. Every response reinforced the boundary.
DESIGN ADVISOR
Prof. Shah Noor Shafqa
Every pixel of whitespace is a moment to breathe
Reframed aesthetics as clinical care. On Day 4, a calm interface isn't preference, it's medicine. Whitespace became a design principle.
DESIGN ADVISOR
Prof. Shah Noor Shafqa
Every pixel of whitespace is a moment to breathe
Reframed aesthetics as clinical care. On Day 4, a calm interface isn't preference, it's medicine. Whitespace became a design principle.
02 - What the literature said
Three findings reshaped the direction
Three findings reshaped the direction
drag me
KEY TAKEAWAYS FROM 11 PAPERS
KEY TAKEAWAYS FROM 11 PAPERS
a. Content must adapt to the individual, literacy alone doesn't predict adherence.
Content must adapt to the individual, literacy alone doesn't predict adherence.
Ha et al. 2022
Ha et al. 2022
b. Multimodal delivery (text + audio + video) triples engagement over text-only.
b. Multimodal delivery (text + audio + video) triples engagement over text-only.
Shaffer et al. 2023
Shaffer et al. 2023
c. Chatbots can outperform nurse-led education, but only with clinical guardrails.
c. Chatbots can outperform nurse-led education, but only with clinical guardrails.
Tawfik 2023 & Lawson McLean 2025
Tawfik 2023 & Lawson McLean 2025
03 - The competitive gap
7 apps · 15 dimensions · 2 buckets, neither of which serves the patient
7 apps · 15 dimensions · 2 buckets, neither of which serves the patient
drag me
INFO-HEAVY APPS
Overwhelming
Static content. No real-time support.
INFO-HEAVY APPS
Overwhelming
Static content. No real-time support.
INFO-HEAVY APPS
Overwhelming
Static content. No real-time support.
CHEMOBUDDY
The middle ground
Education, tracking, and real-time together.
CHEMOBUDDY
The middle ground
Education, tracking, and real-time together.
CHEMOBUDDY
The middle ground
Education, tracking, and real-time together.
TASK-ONLY TRACKERS
Narrow
Log-only. No education layer.
TASK-ONLY TRACKERS
Narrow
Log-only. No education layer.
TASK-ONLY TRACKERS
Narrow
Log-only. No education layer.
04 - Role-play testing
No active patients within timeline. Clinician-guided simulation instead.
No active patients within timeline. Clinician-guided simulation instead.
THE METHOD
Moderator
Dr. Umar
Persona
67 y/o, low tech literacy
Protocol
Think-aloud
Rounds
3 × 15 min
THE METHOD
Moderator
Dr. Umar
Persona
67 y/o, low tech literacy
Protocol
Think-aloud
Rounds
3 × 15 min
" I don't need more information. I need to know. if this is normal right now. "
- TEAM MEMBER, IN ROLE AS A 67-YEAR-OLD PATIENT
" I don't need more information. I need to know. if this is normal right now. "
- TEAM MEMBER, IN ROLE AS A 67-YEAR-OLD PATIENT
That reframed the strategy - from "here's everything" to "here's what's next."
That reframed the strategy - from "here's everything" to "here's what's next."
05 - Meet Sarah
Sarah L. - 58, Stage II breast cancer, Day 4 of her 3rd chemo cycle. Suburban, lives with spouse, moderate tech literacy, prefers audio and visual content over dense text.
Sarah L. - 58, Stage II breast cancer, Day 4 of her 3rd chemo cycle. Suburban, lives with spouse, moderate tech literacy, prefers audio and visual content over dense text.
"On Day 4 of my 3rd chemo cycle, I feel so exhausted that even making tea feels overwhelming. I don't know if this fatigue is normal or if I should call my doctor."
"On Day 4 of my 3rd chemo cycle, I feel so exhausted that even making tea feels overwhelming. I don't know if this fatigue is normal or if I should call my doctor."
"Would this help Sarah on Day 4?"
"Would this help Sarah on Day 4?"
drag me
PRODUCT
PRODUCT
The final app is built around three ideas.
Each solves a different piece of the same problem. Together, they behave like one tool, not three features stitched into an app.
Each solves a different piece of the same problem. Together, they behave like one tool, not three features stitched into an app.
1.
A 3D Symptom Tracker that speaks without words.
Tap where it hurts on a body map, no medical vocabulary needed. Three zones expand into sub-regions with icon-based symptoms. When severity crosses red-flag territory, the system escalates immediately.
Tap where it hurts on a body map, no medical vocabulary needed. Three zones expand into sub-regions with icon-based symptoms. When severity crosses red-flag territory, the system escalates immediately.
2.
An Education Hub that meets patients where their energy is.
Every answer surfaces four formats: Video · Audio · Article · Simplify. Content adapts to the treatment phase and reported symptoms. On a low-energy day, a patient can listen instead of read.
Every answer surfaces four formats: Video · Audio · Article · Simplify. Content adapts to the treatment phase and reported symptoms. On a low-energy day, a patient can listen instead of read.
3.
A Dashboard that makes low-energy days navigable.
One screen replaces sticky notes and five apps: EHR-synced appointments, visual pill photos (not just names), care team contacts, and a "Today's Focus" that surfaces only what matters right now.
One screen replaces sticky notes and five apps: EHR-synced appointments, visual pill photos (not just names), care team contacts, and a "Today's Focus" that surfaces only what matters right now.
Visuals of ChemoBuddy
Flow 1: Onboarding
Goal
Get a newly diagnosed patient from download to their first useful screen in under 60 seconds, before anxiety sets in.
Get a newly diagnosed patient from download to their first useful screen in under 60 seconds, before anxiety sets in.
What it does
•
•
Clinic code login: one input connects the patient's treatment data securely through the hospital's API. No forms, no friction.
Clinic code login: one input connects the patient's treatment data securely through the hospital's API. No forms, no friction.
•
•
Automatic EHR sync: chemotherapy schedules, medications, and clinician notes import in the background. Zero manual entry.
Automatic EHR sync: chemotherapy schedules, medications, and clinician notes import in the background. Zero manual entry.
•
•
Progressive disclosure: the onboarding introduces one feature at a time, dashboard, chatbot and care team, so nothing overwhelms on day one.
Progressive disclosure: the onboarding introduces one feature at a time, dashboard, chatbot and care team, so nothing overwhelms on day one.
Flow 2: Chatbot: Ask ChemoBuddy
Goal
Answer the questions that come at 2 AM. Safe, clinical, and always clear about where education ends and a doctor begins.
Answer the questions that come at 2 AM. Safe, clinical, and always clear about where education ends and a doctor begins.
What it does
•
•
Context-aware responses: recognizes treatment phase and symptom severity to tailor tone and depth. Day 1 answers differ from Day 4 answers.
Context-aware responses: recognizes treatment phase and symptom severity to tailor tone and depth. Day 1 answers differ from Day 4 answers.
•
•
Multimodal education: after every response, patients choose their format, video, audio, article, or simplified text. Energy decides, not the developer.
Multimodal education: after every response, patients choose their format, video, audio, article, or simplified text. Energy decides, not the developer.
•
•
Red-flag escalation: NLP detects critical keywords (fever ≥104°F, chest pain, breathing difficulty) and interrupts with an emergency alert + one-tap [Call Clinic] / [Find ER].
Red-flag escalation: NLP detects critical keywords (fever ≥104°F, chest pain, breathing difficulty) and interrupts with an emergency alert + one-tap [Call Clinic] / [Find ER].
Flow 3: Dashboard
Goal
A calm home base for low-energy days, treatment progress, medications, and what matters today in one glance.
A calm home base for low-energy days, treatment progress, medications, and what matters today in one glance.
What it does
•
•
Treatment progress ring: circular visualization showing completed and upcoming cycles, with milestone messaging that evolves as the patient advances.
Treatment progress ring: circular visualization showing completed and upcoming cycles, with milestone messaging that evolves as the patient advances.
•
•
Today's Focus: surfaces only what matters right now, next appointment, current medication, and hydration reminder. Nothing else competes for attention.
Today's Focus: surfaces only what matters right now, next appointment, current medication, and hydration reminder. Nothing else competes for attention.
Flow 4: 3D Symptom Tracker
Goal
Let patients show where it hurts without knowing the medical word for it.
Let patients show where it hurts without knowing the medical word for it.
What it does
•
•
Interactive body map: three zones (Head & Neck, Upper Body, and Lower Body) expand into sub-regions. Patients tap, not type. No medical vocabulary required.
Interactive body map: three zones (Head & Neck, Upper Body, and Lower Body) expand into sub-regions. Patients tap, not type. No medical vocabulary required.
•
•
Severity and type selection: icon-based symptom categories (pain, nausea, burning, numbness, swelling) with a 1–5 severity scale. Color-coded feedback shows green to red.
Severity and type selection: icon-based symptom categories (pain, nausea, burning, numbness, swelling) with a 1–5 severity scale. Color-coded feedback shows green to red.
•
•
Automatic escalation: when symptoms cross into dangerous territory, the system doesn't wait for the patient to recognize it, it flags the care team immediately.
Automatic escalation: when symptoms cross into dangerous territory, the system doesn't wait for the patient to recognize it, it flags the care team immediately.
TESTING
TESTING
Three iterations. Three changes that mattered.
Moderated think-aloud sessions, three rounds of 15 minutes each. Dr. Umar evaluated clinical safety and AI boundaries while team members role-played Sarah (the patient) and her daughter (the caregiver) using validated personas.
Moderated think-aloud sessions, three rounds of 15 minutes each. Dr. Umar evaluated clinical safety and AI boundaries while team members role-played Sarah (the patient) and her daughter (the caregiver) using validated personas.
STEPS
STEPS
4 task scenarios
3 iterations
Clinician-moderated
4 task scenarios
Clinician-moderated
3 iterations
TASK
TASK
Iteration 1 - Red-Flag Safety Protocol
Problem
"I have a 104° fever" got an article back. No alert. No safety net.
Design move
NLP detection → "Attention Needed!" alert + one-tap [Call Clinic] / [Nearest ER].
Outcome
100% critical keywords detected.
Iteration 1 - Red-Flag Safety Protocol
Problem
"I have a 104° fever" got an article back. No alert. No safety net.
Design move
NLP detection → "Attention Needed!" alert + one-tap [Call Clinic] / [Nearest ER].
Outcome
100% critical keywords detected.
Iteration 2 - The Educational Basket
Problem
Text-only chatbot. Day 4 patients were too tired to read.
Design move
Every answer ends with a format choice: Video session · Audio explanation · Flashcards · Text guide.
Outcome
85% engagement lift (4.2 → 7.8 min).
Iteration 2 - The Educational Basket
Problem
Text-only chatbot. Day 4 patients were too tired to read.
Design move
Every answer ends with a format choice: Video session · Audio explanation · Flashcards · Text guide.
Outcome
85% engagement lift (4.2 → 7.8 min).
Iteration 3 - Emotional Milestones
Problem
Progress: 10%" felt like 90% left to survive.
Design move
Same number. New meaning: "You've made a strong start. Keep going."
Outcome
Closes the survivorship gap, no existing tool addresses it.
Iteration 3 - Emotional Milestones
Problem
Progress: 10%" felt like 90% left to survive.
Design move
Same number. New meaning: "You've made a strong start. Keep going."
Outcome
Closes the survivorship gap, no existing tool addresses it.
IMPACT
IMPACT
The product moved real numbers on both sides.
Outcomes were tested based on each
patient's experience.
Outcomes were tested based on each
patient's experience.
01. PATIENT -
01. PATIENT -
Did the app help?
Did the app help?
95%
task completion (all 4 scenarios completed unassisted)
95%
task completion (all 4 scenarios completed unassisted)
85%
engagement lift (4.2 → 7.8 min sessions)
85%
engagement lift (4.2 → 7.8 min sessions)
30%
improvement in symptom reporting accuracy
30%
improvement in symptom reporting accuracy
02. CLINICAL -
02. CLINICAL -
Was it safe?
Was it safe?
100%
red-flag detection rate on Dr. Umar's validation pass
100%
red-flag detection rate on Dr. Umar's validation pass
Escalation logic reviewed and approved against Mayo's clinical criteria
Escalation logic reviewed and approved against Mayo's clinical criteria
Every session made one thing clear: the chatbot educates, it doesn't diagnose.
Every session made one thing clear: the chatbot educates, it doesn't diagnose.
03. VALIDATION -
03. VALIDATION -
Did the bet pay off?
Did the bet pay off?
1st
Education, symptom tracking and care coordination; no competitor did all three.
1st
Education, symptom tracking and care coordination; no competitor did all three.
Day 4
Persona-driven design ("Day 4 test") caught feature creep before it shipped
Day 4
Persona-driven design ("Day 4 test") caught feature creep before it shipped
3 + 1
Stakeholder validation from three clinical mentors and one design advisor
3 + 1
Stakeholder validation from three clinical mentors and one design advisor
"It doesn't feel like an app, it feels like a companion."
"It doesn't feel like an app, it feels like a companion."
- Test Participant
- Test Participant
LEARNING
LEARNING
What This Project Taught Me
What This Project Taught Me
Lessons learned about user behavior, product strategy, and designing within constraints.
Whitespace is care.
For a patient on Day 4, every pixel of visual noise adds stress. Every pixel of breathing room removes it. In healthcare, aesthetic decisions aren't cosmetic; they're clinical.
Whitespace is care.
For a patient on Day 4, every pixel of visual noise adds stress. Every pixel of breathing room removes it. In healthcare, aesthetic decisions aren't cosmetic; they're clinical.
The "Day 4 Test" prevents feature creep.
"Would this help Sarah on Day 4?" was asked at every decision. If a feature required focus or reading, it didn't belong in v1. Constraints made the product better.
The "Day 4 Test" prevents feature creep.
"Would this help Sarah on Day 4?" was asked at every decision. If a feature required focus or reading, it didn't belong in v1. Constraints made the product better.
Designing for the hardest user helps everyone.
Large touch targets, audio content, and icon-based interactions are built for a 58-year-old with low literacy and chemo fatigue. Every test participant preferred them. Accessibility isn't an edge case. It's the baseline.
Designing for the hardest user helps everyone.
Large touch targets, audio content, and icon-based interactions are built for a 58-year-old with low literacy and chemo fatigue. Every test participant preferred them. Accessibility isn't an edge case. It's the baseline.
Timing beats thoroughness.
The right answer at 2 AM is worth more than a complete library at 10 AM. Just-in-time guidance outperformed comprehensive resources in every test. Context is the product.
Timing beats thoroughness.
The right answer at 2 AM is worth more than a complete library at 10 AM. Just-in-time guidance outperformed comprehensive resources in every test. Context is the product.
A digital tool must never create false security.
The most important constraint on the whole project. The chatbot educates. It doesn't diagnose. Every response reinforces the boundary. When in doubt, we escalated. Always.
A digital tool must never create false security.
The most important constraint on the whole project. The chatbot educates. It doesn't diagnose. Every response reinforces the boundary. When in doubt, we escalated. Always.
What comes next
What comes next
Each iteration helped the user to build confidence and improve collaboration.
Each iteration helped the user to build confidence and improve collaboration.
Next Steps - short term:
a. IRB testing with real oncology patients and a patient advisory board for content review.
b. TV and desktop parity: older patients find mobile difficult during long infusion sessions.
Next Steps - short term:
a. IRB testing with real oncology patients and a patient advisory board for content review.
b. TV and desktop parity: older patients find mobile difficult during long infusion sessions.
Future Ideas - long term:
a. Clinician-facing dashboard surfacing red-flag alerts and patient-reported outcomes.
b. Multilingual accessibility, Spanish and Mandarin first, with cultural adaptation, not just translation.
Future Ideas - long term:
a. Clinician-facing dashboard surfacing red-flag alerts and patient-reported outcomes.
b. Multilingual accessibility, Spanish and Mandarin first, with cultural adaptation, not just translation.
NEXT PROJECT
NEXT PROJECT
©2026 Rishi Kataria • Arizona
©2026 Rishi Kataria • Arizona
©2026 Rishi Kataria • Arizona