Overview
IOS App Design
| Quantitative User Research
| VUI Design
| AI Agent
| Branding
At UC Berkeley’s entrepreneurship project Harmony, I turned voice-biomarker ML into an empathetic, consent-first VUI with micro-coaching. The system translates signals into actionable, context-aware guidance for middle-aged and older users—boosting daily engagement without medical over-claim.
Type:
Entrepreneurship @UCB
Industry:
HealthTech, AI
Role:
Founding Product Designer
Timeline:
Jan 2024 - Aug 2024
Built a consent-first VUI that turns voice-biomarker ML into actionable micro-coaching—empathetic at home, compliant, and engagement-focused.
From Voice Biomarkers to Daily Health UX
Current Challenges
Dementia screening
Dementia is a disease that affects countless people worldwide, either directly or through a loved one. However, traditionally dementia is diagnosed using PET (Positron Emission Tomography) imaging and cerebrospinal fluid exams to measure the concentration of amyloid plaques in the brain, a costly and invasive process. A more cost-effective, non-invasive and easily-accessible technique is needed.
Opportunities at AI Age
Detecting Dementia just using human's voice
At UC Berkeley, I formed an Interdisciplinary & International Team Harmony at SCET as the sole designer. We have successfully trained 2 AI models to proactively predict dementia through voice biomarkers, reached over 80% accuracy.
Model 1: Regression
Techniques: XGBoost, KNN, Random Forest, VotingClassifier, Standard Scalar, Gaussian Weights, CrossValidation
Model 2: Neural Network
Techniques: LSTM, Sigmoid, GeLU, Dense, Dropout,Adam, SelectKBest, Gradient Clipping, Standard Scalar, EarlyStopping
Design Objectives
Link emerging tech to real-world users
In this project, we take on the role of entrepreneurs to explore how cutting-edge technology can be seamlessly integrated into real users' lives. As the sore designer on team, I turned voice-biomarker ML into an empathetic, context-aware VUI that supports natural, compassionate home interactions—bridging cutting-edge AI and human-centered care.
Immerse to Understand Tech
Neurotech exploration hackathon
Key Findings:
User Interview & Analysis
Discover and engage with early customers
3
4
5
2
1
0%
0%
10%
10%
10%
10%
20%
20%
20%
30%
20%
30%
30%
40%
30%
40%
40%
50%
40%
50%
50%
50%
60%
60%
60%
70%
60%
70%
70%
80%
70%
80%
80%
80%
11.5%
5.5%
2.7%
15.2%
23.3%
23.1%
56%
82%
19.2%
15.4%
16%
12%
12%
14.7%
14.0%
11.0%
16.3%
34.6%
4%
9.7%
7.6%
34.8%
14.0%
49.0%
Likert Scale
customers’ interests, trusts and willingness?
I am willing to use Harmony for periodic tracking.
I am willing to share voice data with Harmony to track.
I am willing to trust Harmony’s predictions.
I am willing to use it to monitor for dementia prevention.
Open Questions:
Likert Scale Questions (requires rating from 1-5):
How do you think of the idea?
... ...
Do you think this product should be more medical/serious/efficient or more interactive/positive/friendly?
We interviewed 72 individuals age 45-65 due to their higher risk of cognitive decline.
Semi-structured
72 Participants
Berkeley, CA
4 Weeks
Key findings:
1. People are intrigued and amazed by this convenient technology but remain skeptical about AI.
2.People believe that intuitive and friendly interfaces outweigh complex and technical ones.
3.Age-based customer division is too vague!
User Interview & Analysis
Cross-stakeholder Interviews
Key findings:
1. AI explainability is key to building user trust.
2.Partnering with medical platforms for endorsement enhances credibility.
3.Combining useful advice with detection is essential, as pure monitoring alone won't keep users engaged.
Define Persona & User Journey Map
Segmented
3
4
5
2
1
0%
10%
10%
20%
20%
30%
30%
40%
40%
50%
50%
60%
70%
90%
80%
100%
When compared to being filtered earlier, positivity (5 points) has increased.
The vast majority are very acceptable to use it long-term.
Clara,45
highly focused on monitoring her cognitive and physical health due to hereditary risks.
Persona 1
User Persona
Unlike physical diagnosis, cognitive diagnosis should base on long term. However, under most situations, doctors don’t know their patients well.
Some tests are invasive.
“Am I just over-concerned? Whether it is just a normal thing when aging?”
“Being supervised is stressed”
Fill classical measure test under monitor pressure
Doctors lack long term data of patients
Long journey before
having an answer
Hard to track consecutively
Unsure about whether going to have a formal test at hospital
3
4
5
6
2
Identify Pain Points from User Journey
We re-filtered the data and found that, within a specific user group, the indicators were highly promising.
Close Relatives of Dementia Patients
Likert Scale
Segmented Customers’
According to the Alzheimer's Association, About 30-50% of Alzheimer's patients have a family history, raising their risk by 2-4 times.
Family history raises Alzheimer's risk by 2-4 times.
often feel tired and become forgetful
Notice typical symptoms
with suspect
1
Ribecca,62
prioritizes health and stays proactive about maintaining cognitive well-being as ages.
Proactive Cognitive Health Seeker
Persona 2
Key findings:
1. AI explainability is key to building user trust
2.Partnering with medical platforms for endorsement enhances credibility.
3.Combining useful advice with detection is essential, as pure monitoring alone won't keep users engaged.
Marketing Research
Identify challenges and opportunities in market
Competitive Canvas
Market Size
$5.6 Trillion
$29 Billion
Health & Wellness
Ecosystem
Dementia
Care
Total Available Market: $3.9B
11 million family caretakers
Serviceable Available Market: $59M
1.5% Market Acquisition
Serviceable btainable Market: $5M
10% capture of SAM
Product Goal:
A proactive, regular, at-home early detection
Based on user feedback, we aim to deliver an integrated solution powered by our AI ML model that connects high-risk dementia individuals, their families, and medical platforms. This solution leverages voice interaction and family sharing features to seamlessly integrate the technology into users' daily lives.
MVP Design
Using wireframe to rapidly create MVP for testing and iteration
Open-Testing
Open Testing @Berkeley The Large Bid
Testing Setup:
The Large Bid is a public startup pitch event held at UC Berkeley's SCET. At the event, we invited 12 participants to test our prototype. The testers included students with family members suffering from dementia, as well as middle-aged investors, among others.
Research Poster
Back-end Server
Test Device
Testing Outline
Key Findings:
Final Layout & Iteration
Smart Topic
Obtain Apple Health permissions and tailor health reflection topics based on the user's activity and usage patterns.
Responsive Dialogue Interaction
As a startup originating from UC Berkeley, we began with a tangible conversational agent inspired by the university's mascot bear, featuring rich expressions and movements. Beyond of this defalut character, we also allow users to customze their chatting partner. This creates a warm, engaging, and interactive communication experience that boosts user engagement and retention.
Family-sharing & Privacy Protection
Family-sharing: We also addressed key needs and concerns related to family sharing and data privacy mentioned in user research. By offering users the option to choose between direct users and caregivers, we segmented the monitoring experience to meet the core needs of different stakeholders.
Privacy Protection: We emphasized the protection of voice data and clearly defined the scope of its application.
Daily Report
Summary page: Users will receive a daily report after completing their voice input, including basic cognitive analysis, improvement suggestions, and a log of the day’s conversation.
Voice Biomarker Analysis: Explain the parameter meanings and optimal ranges, emphasizing education and scientific accuracy.
Cognitive Tendency Report: After reaching the required interactions, Harmony will analyze cognitive health trends and offer tailored recommendations.
“Where PS meets Figma” in BioTech @ Cellverse AI
B2B | Web UIUX | Data Visualization
Product Innovation @Siemens Healthineers
B2B | Strategy Design | Product Design | Case Study
B2C | UX Research | User Interview | Insights Alignment
UXR & Inno-Design
@Siemens Healthineers
Field Visit | Contextual Inquiry | Insights Alignment| Concept Design