Overview
Neuropy is a mood-tracking system that combines a conversational AI Home Hub with passive physiological data to enable continuous, low-effort mood logging. It improves tracking consistency by 32% and provides clinicians with actionable insights to support more informed treatment decisions.
My Role
Sole Product Designer
UX Researcher
Team
1 Project Manager
1 Software Engineer
1 Industrial Designer
Timeline
Sep 2024 - March 2025
A quick snapshot of what we've built
Why mood tracking is important?
We interviewed mental health clinicians, scholar, five potential users to understand how mood disorders are treated and what we can do to help with mental health management.
Current mood-tracking methods meet some challenges
There are two common ways that people track their mood, one is self-assessment questionnaires like PHQ-9 and GAD-7, and the other is mood-tracking app, such as Daylio and Finch, which are the two frequently-use apps in our user interview.
Mood-tracking retention drops from 12% to 5.7% in 30 days, as manual input quickly becomes unsustainable.
Existing apps offer limited emotion options, making it hard for users to express nuanced feelings accurately.
Apps show mood history but fail to link emotions to lifestyle factors, leaving users unsure how to improve.
Complex forms and frequent notifications overwhelm users, discouraging consistent mood tracking.
After ideation and extensive research , we decided to make mood logging automatically by using physiological data to estimate user’s mood state.
Wearable Device
Collect Biometric Data
Acquire Data by API
Estimate Mood by ML model
Visualize Mood Result on Phone
We built a prototype to visualize the mood estimation result.
However, user testing challenged our initial hypothesis.
Through user testing, we found that users didn’t trust our mood-estimation approach. Instead, they concerns around trust, meaning, and emotional connection:
Learning from users, we took a step back and reframed our focus.
With the new design focus, we went through several rounds ideation and validation.
To ensure reliable mood insights, we proposed a continuous mood-tracking model by combining three data sources.
Physiological Data
An AI-driven algorithm estimates the user's mood by analyzing biometric data.
Speech Data
An AI-powered Home Hub analyzes the user's conversational data, capturing emotional nuances and contextual meaning.
Feature #1
Home Hub captures emotional context and nuances.
Our Home Hub is designed to capture complex emotions by analyzing speech data, users can log their mood by simply talking about their day, and they can get the emotional summary on the mobile app.
9:41
Mood Report
Daily Topic
Summary
(generated by AI)
Mood Breakdown
(understand proportion)
Mood Triggers
(categorized by factors, e.g. location, people, event)
Different tone with the same words can lead to different meanings. We connect our Home Hub with Hume.ai API to further analyze user’s voice expression.
We designed the Mood Cloud to visualize how users’ emotions evolve over time.
Mood cloud will evolve over time as each different emotions detected. Users can also view it by the “factor” category, helping them identify mood triggers in an interactive way.
9:41
9:41
9:41
9:41
This video shows how the Home Hub actually works
Feature #2
Mood dashboard presents mood data points.
Three data sources were combined to display in one place for better comparison
Mood Journal & Supplementary Info
(Summary from Home Hub data)
We iterated the mood visualization to better represent multiple types of emotional data.
Version 1
Average mood score line
Line graph for adjustments
Version 2
Emotion-specific markers
Continuous mood points
Version 3
Stacked bar graph
Discrete mood points
#1: From Numbers to Emotion Labels
Single mood number oversimplified emotional states. Users wanted to see more detailed mood fluctuations across different emotions.
#2: Expand Multi-Emotion Comparison
Mood states on different days are distinct and multi-layered. Users wanted a clearer way to compare multiple emotions over time.
We also correlated mood patterns with lifestyle factors for better management
Feature #3
Theory-based mood wheel for logical mood logging.
To parse complex emotions in a logical and scientific way, we applied Plutchik’s Circumplex Model of Emotions to design the mood wheel, helping users identify eight primary emotions, understand intensity levels, and visualize emotional combinations.
8 Primary Emotions
5 Intensity Levels
9:41
We created playful emotion characters to represent the eight core moods, making mood logging more engaging and interactive.
some emojis credit to Raz Rashid
Behind the scene...
Over six months, we progressed from problem definition and concept development to prototyping and technical implementation. The process was grounded in solid user research and iterative design, making the system both comprehensive and expert-backed.
What I Learned...
Early assumptions about automation were overturned by real user feedback, reminding me I should be open to pivoting when research reveals deeper user needs.
Multiple rounds of iteration to simplify complex emotional information taught me that a good visualization isn't about showing everything—it's about showing what matters.
Designing for mental health reminded me that UX isn’t just about usability; it’s about emotional resonance. Tone, wording, and emotional context all shape trust, comfort, and engagement.
