From Data Chaos to Clarity

AI-powered tools for faster, clearer decisions

THE PROJECT

Overview

CuraJOY is a nonprofit tech company making behavioral and mental health support more accessible through gamification, self check-ins, and AI-powered tools.


In behavioral health, Board Certified Behavior Analysts (BCBAs) often have just 6–12 billable hours to complete a Functional Behavior Assessment (FBA)—far less than the time needed to:

CuraJOY is a nonprofit tech company making behavioral and mental health support more accessible through gamification, self check-ins, and AI-powered tools.


In behavioral health, Board Certified Behavior Analysts (BCBAs) often have just 6–12 billable hours to complete a Functional Behavior Assessment (FBA)—far less than the time needed to:

CuraJOY is a nonprofit tech company making behavioral and mental health support more accessible through gamification, self check-ins, and AI-powered tools.


In behavioral health, Board Certified Behavior Analysts (BCBAs) often have just 6–12 billable hours to complete a Functional Behavior Assessment (FBA)—far less than the time needed to:

  • Conduct caregiver and staff interviews

  • Observe clients across home, school, and clinic settings

  • Collect and analyze varied data sources (notes, reports, videos)
    Identify behavior functions and draft documentation

This process is fragmented and labor-intensive. MyCuraJOY’s B2B platform aims to fix that with automated behavior labeling, trend summaries, visualizations, and functional insights—reducing time and improving accuracy.

This process is fragmented and labor-intensive. MyCuraJOY’s B2B platform aims to fix that with automated behavior labeling, trend summaries, visualizations, and functional insights—reducing time and improving accuracy.

This process is fragmented and labor-intensive. MyCuraJOY’s B2B platform aims to fix that with automated behavior labeling, trend summaries, visualizations, and functional insights—reducing time and improving accuracy.

My Role

Design Lead

Led the design efforts across product research, user interviews, user flows, rapid prototyping, and usability testing

Team

Engineering Lead

Billy Franklin

Researchers

Haley Scheer; Jessica Peng; Laura Heppell

Timeline

Feb 2025 - Jul 2025

Results & Impact

62%

62%

Reduction in behavior plan prep time (cut from 1.5 hours to 45 minutes) through automated data labeling and graph generation

90%+

90%+

AI label accuracy sustained with human-in-the-loop reviews

38%

38%

Boost in analyst satisfaction after cutting manual data entry and unifying tools into a single workflow

Auto Behavior Summary

Dynamic Interactive
Behavior Graph

Dynamic Interactive
Behavior Graph

Context-Aware AI Chat

Context-Aware
AI Chat

REQUIRNMENTS

Project Scope

The backend AI already labels behavior logs with over 90% accuracy, but BCBAs still need an efficient way to analyze those labels.

My scope:


  • Design tools that visualize and summarize behavior patterns

  • Create graphs BCBAs can easily include in behavior intervention plans (BIP)

My scope:


  • Design tools that visualize and summarize behavior patterns

  • Create graphs BCBAs can easily include in behavior intervention plans (BIP)

My scope:


  • Design tools that visualize and summarize behavior patterns

  • Create graphs BCBAs can easily include in behavior intervention plans (BIP)

RESEARCH

Discovering Pain Points

Our research team interviewed eight behavior analysts and identified four key pain points:

Our research team interviewed eight behavior analysts and identified four key pain points:

Our research team interviewed eight behavior analysts and identified four key pain points:

  1. Unclear ABC Data – Hard to identify actionable patterns; reviewing logs is time-consuming.


  2. Inconsistent Data Collection – Human error, distractions, and limited training reduce reliability.


  3. Missing Environmental Context – Key situational factors are often absent or poorly documented.


  4. Documentation Overload – Large volumes of unstructured data increase workload instead of streamlining it.

  1. Unclear ABC Data – Hard to identify actionable patterns; reviewing logs is time-consuming.


  2. Inconsistent Data Collection – Human error, distractions, and limited training reduce reliability.


  3. Missing Environmental Context – Key situational factors are often absent or poorly documented.


  4. Documentation Overload – Large volumes of unstructured data increase workload instead of streamlining it.

  1. Unclear ABC Data – Hard to identify actionable patterns; reviewing logs is time-consuming.


  2. Inconsistent Data Collection – Human error, distractions, and limited training reduce reliability.


  3. Missing Environmental Context – Key situational factors are often absent or poorly documented.


  4. Documentation Overload – Large volumes of unstructured data increase workload instead of streamlining it.

When I joined the project, I conducted follow-up interviews with two curaJOY’s BCBA consultants and learned that:

Most BCBAs track no more than 3 behaviors per billable period

Most BCBAs track no more than 3 behaviors per billable period

Behavior-over-time graphs are especially valuable during the FBA process

Most BCBAs track no more than 3 behaviors per billable period

Behavior-over-time graphs are especially valuable during the FBA process

Behavior-over-time graphs are especially valuable during the FBA process

Quickly spotting trends is critical for making efficient and effective decisions

Quickly spotting trends is critical for making efficient and effective decisions

Quickly spotting trends is critical for making efficient and effective decisions

User Persona

To stay focused on the core needs of BCBAs and help stakeholders quickly understand the main user group’s pain points, I developed this persona:

Dr. Mandy

BCBA-D

Prioritizes accuracy, efficiency, and clear insights in her workflow

Behavior

  • Oversees assessments and treatment plans for multiple clients

  • Reviews behavior data (ABC logs) collected by teachers and caregivers

  • Creates visual reports for FBAs and BIPs

Pain Points

  • Struggles to extract actionable insights from lengthy, unstructured ABC logs

  • Faces inconsistencies in data due to human error and lack of training in data entry

  • Lacks adequate documentation of environmental factors influencing behaviors

  • Spends significant time organizing and synthesizing raw data into usable formats

  • Struggles to extract actionable insights from lengthy, unstructured ABC logs

  • Faces inconsistencies in data due to human error and lack of training in data entry

  • Lacks adequate documentation of environmental factors influencing behaviors

  • Spends significant time organizing and synthesizing raw data into usable formats

  • Struggles to extract actionable insights from lengthy, unstructured ABC logs

  • Faces inconsistencies in data due to human error and lack of training in data entry

  • Lacks adequate documentation of environmental factors influencing behaviors

  • Spends significant time organizing and synthesizing raw data into usable formats

Core Needs

  • Tools that clearly highlight behavior trends over time

  • Ability to capture and integrate environmental context into reports

  • Streamlined reporting features to reduce documentation workload

  • Tools that clearly highlight behavior trends over time

  • Ability to capture and integrate environmental context into reports

  • Streamlined reporting features to reduce documentation workload

  • Tools that clearly highlight behavior trends over time

  • Ability to capture and integrate environmental context into reports

  • Streamlined reporting features to reduce documentation workload

RESEARCH

Insights From Existing Tools

To gather inspiration, I reviewed tools aligned with the project’s focus areas:

AI Tools: ChatGPT, Gemini, Claude

Graphing Tools: Tableau, Power BI

AI Tools: ChatGPT, Gemini, Claude

Graphing Tools: Tableau, Power BI

AI Tools: ChatGPT, Gemini, Claude

Graphing Tools: Tableau, Power BI

RESEARCH

AI Tools

Strengths

  • Can quickly summarize large datasets

  • Flexible natural language interaction

  • Integration-ready via APIs

Gaps

  • Limited domain-specific accuracy without custom training

  • No built-in FBA/BIP behavioral health frameworks

  • Lack of persistent memory across sessions for ongoing cases

Graphing Tools

Strengths

  • Advanced filtering, drill-down, and visualization options

  • Supports multiple data source integrations

  • Strong export and sharing capabilities

  • Customizable dashboards for different audiences

Gaps

  • Steep learning curve for non-data specialists

  • Requires manual data preparation and cleaning

  • No built-in contextual or narrative explanations for graphs

How might we use AI to turn messy behavior logs into clear, actionable insights—fast?

DESIGN

Smart Tools for Complex Cases

From both team research and my follow-up interviews, I learned that every client’s FBA case is unique. Some behavior patterns can be identified quickly through frequency or duration, while others require deeper analysis across multiple settings—beyond just school or home.

From both team research and my follow-up interviews, I learned that every client’s FBA case is unique. Some behavior patterns can be identified quickly through frequency or duration, while others require deeper analysis across multiple settings—beyond just school or home.

From both team research and my follow-up interviews, I learned that every client’s FBA case is unique. Some behavior patterns can be identified quickly through frequency or duration, while others require deeper analysis across multiple settings—beyond just school or home.

To support this variability without overwhelming users, I started with a static Figma prototype showing AI chat flows and simple graphs to validate the concept. After reviewing feasibility with our Engineering Lead and gathering input from BCBA consultants, I built an interactive prototype using Loveable—integrating ChatGPT 4.0 for conversations and ClaudeCode for dynamic graphing.


The prototypes were guided by these core principles:


  • Provide quick, clear summaries of behavior trends.

  • Ensure transparency so BCBAs can understand and trust AI analysis.

  • Offer nudges to help BCBAs decide next steps without feeling overwhelmed.

  • Enable customizable, filterable graphs that let BCBAs explore data points, consequences, and setting correlations with the behaviors they’re targeting.

To support this variability without overwhelming users, I started with a static Figma prototype showing AI chat flows and simple graphs to validate the concept. After reviewing feasibility with our Engineering Lead and gathering input from BCBA consultants, I built an interactive prototype using Loveable—integrating ChatGPT 4.0 for conversations and ClaudeCode for dynamic graphing.


The prototypes were guided by these core principles:


  • Provide quick, clear summaries of behavior trends.

  • Ensure transparency so BCBAs can understand and trust AI analysis.

  • Offer nudges to help BCBAs decide next steps without feeling overwhelmed.

  • Enable customizable, filterable graphs that let BCBAs explore data points, consequences, and setting correlations with the behaviors they’re targeting.

To support this variability without overwhelming users, I started with a static Figma prototype showing AI chat flows and simple graphs to validate the concept. After reviewing feasibility with our Engineering Lead and gathering input from BCBA consultants, I built an interactive prototype using Loveable—integrating ChatGPT 4.0 for conversations and ClaudeCode for dynamic graphing.


The prototypes were guided by these core principles:


  • Provide quick, clear summaries of behavior trends.

  • Ensure transparency so BCBAs can understand and trust AI analysis.

  • Offer nudges to help BCBAs decide next steps without feeling overwhelmed.

  • Enable customizable, filterable graphs that let BCBAs explore data points, consequences, and setting correlations with the behaviors they’re targeting.

Design Evolutions

AI Analysis

Dynamic Graph

USABILITY TESTING

Testing Method

Method

Moderated 30 minutes user usability testing

Participants

10 BCBAs who regularly conduct FBA and BIP

All BCBAs in the study had recent hands-on experience with FBAs and BIPs and passed screening questions to confirm they understood graphing and functional labels. Each joined me on a Zoom call to test the interactive prototype in Loveable, where they completed three tasks:

All BCBAs in the study had recent hands-on experience with FBAs and BIPs and passed screening questions to confirm they understood graphing and functional labels. Each joined me on a Zoom call to test the interactive prototype in Loveable, where they completed three tasks:

All BCBAs in the study had recent hands-on experience with FBAs and BIPs and passed screening questions to confirm they understood graphing and functional labels. Each joined me on a Zoom call to test the interactive prototype in Loveable, where they completed three tasks:

Task 1: Select Data Sources – Choose the sources to include in an AI prompt.

Task 1: Select Data Sources – Choose the sources to include in an AI prompt.

Task 1: Select Data Sources – Choose the sources to include in an AI prompt.

Task 2: Generate Behavior Graph – Create a behavior trend graph based on the selected sources.

Task 2: Generate Behavior Graph – Create a behavior trend graph based on the selected sources.

Task 2: Generate Behavior Graph – Create a behavior trend graph based on the selected sources.

Task 3: Identify New Behaviors – Review and flag behaviors appearing outside established trends.

Task 3: Identify New Behaviors – Review and flag behaviors appearing outside established trends.

Task 3: Identify New Behaviors – Review and flag behaviors appearing outside established trends.

Test Results

What Worked

  • AI Analysis: Helpful summaries with selectable sources

  • AI Analysis: Helpful summaries with selectable sources

  • AI Analysis: Helpful summaries with selectable sources

  • Graphs: Heatmap useful for FBAs; filtering by environment/date was valuable

  • Graphs: Heatmap useful for FBAs; filtering by environment/date was valuable

  • Graphs: Heatmap useful for FBAs; filtering by environment/date was valuable

Needs Improvement

  • AI Analysis: Expandable source window; always include comparison context in summaries

  • AI Analysis: Expandable source window; always include comparison context in summaries

  • AI Analysis: Expandable source window; always include comparison context in summaries

  • Graphs: Make line, bar and duration graphs should be the priority

  • Graphs: Make line, bar and duration graphs should be the priority

  • Graphs: Make line, bar and duration graphs should be the priority

User Quotes

"Can you filter by that? Like, the filter by submitter? That's a big thing I liked to do when I look through data…"

"The occurrences are compared to what period of time?…I think I need to see more context"

"We most definitely use line graphs, bar graphs and it is important to see duration as well…"

"…So this attaches the sources to the chat, I think it's great to give it that context"

"I like what I am looking at, very impressed!"

DESIGN ITERATION

User-Informed Redesign

Based on user feedback, I refined key features to better meet their needs. After two rounds of critiques, here are some before-and-after redesigns.

  • Before

    Source window difficult to read; no expansion or filtering

    After

    Added expandable, filterable source column

  • Before

    Behavior summaries lacked contextual comparisons

    After

    Auto summaries include weekly comparison context

  • Before

    Graphs had limited interactivity; no cause/setting correlation view

    After

    Responsive, customizable graphs with relational bar views to link causes/settings to top behaviors for faster insights

WHAT I LEARNED

Small Changes, Big Impact

The prototype is still in development, but this project pushed me to explore new tools and build a fully interactive version. Conversations with BCBAs showed me that small tweaks—like clearer data sources or simpler graph controls—can make a big difference in their daily work and strengthen their trusts in the capabilities of new AI tools. Real feedback proved far more valuable than assumptions, and rapid iterations kept the design practical, trustworthy, and even enjoyable to use.

The prototype is still in development, but this project pushed me to explore new tools and build a fully interactive version. Conversations with BCBAs showed me that small tweaks—like clearer data sources or simpler graph controls—can make a big difference in their daily work and strengthen their trusts in the capabilities of new AI tools. Real feedback proved far more valuable than assumptions, and rapid iterations kept the design practical, trustworthy, and even enjoyable to use.

The prototype is still in development, but this project pushed me to explore new tools and build a fully interactive version. Conversations with BCBAs showed me that small tweaks—like clearer data sources or simpler graph controls—can make a big difference in their daily work and strengthen their trusts in the capabilities of new AI tools. Real feedback proved far more valuable than assumptions, and rapid iterations kept the design practical, trustworthy, and even enjoyable to use.

View More Work...

B2B

SaaS

Route Logistics

Tech For Safer Streets

AI

B2B

SaaS

Behavioral Health

Training AI Through Teamwork

B2C

Wellness App

Behavioral Health

Behavior, Simplified

View More Work...

B2B

SaaS

Route Logistics

Tech For Safer Streets

AI

B2B

SaaS

Behavioral Health

Training AI Through Teamwork

B2C

Wellness App

Behavioral Health

Behavior, Simplified

View More Work...

B2B

SaaS

Route Logistics

Tech For Safer Streets

AI

B2B

SaaS

Behavioral Health

Training AI Through Teamwork

B2C

Wellness App

Behavioral Health

Behavior, Simplified

View More Work...

B2B

SaaS

Route Logistics

Tech For Safer Streets

AI

B2B

SaaS

Behavioral Health

Training AI Through Teamwork

B2C

Wellness App

Behavioral Health

Behavior, Simplified

View More Work...

B2B

SaaS

Route Logistics

Tech For Safer Streets

AI

B2B

SaaS

Behavioral Health

Training AI Through Teamwork

B2C

Wellness App

Behavioral Health

Behavior, Simplified