Fitness Furniture
UX Research
Overview
A generative system using p5.js, LLMs, and parsed policy + heritage archive data to visualize the tension between lived and legislated definitions of culture through Tanzanian Kitenge prints.
Timeline
Ongoing (Prototype completed)
My Role
Creative Technologist & Critical Systems Designer
Tools
p5.js, HTML/CSS, JavaScript, OpenCV, Python, MongoDB, Firebase, Large Language Models (LLMs), Node.js
Teammates
AI/ML Engineers ②
Data Visualization Specialist ①, Cultural Anthropologists ②
Policy Analysts ②
WHAT IS A KITENGE?
Critical analysis of culture
INSPIRATION
THE BIG IDEA
MID-FIDELITY PROTOTYPE
Building the system
REFLECTIONS AND NEXT STEPS
This project began with a deep dive into the functional roles of culture on a governing level. My case study was Tanzania and the role that culture has in shaping our laws and policies. From this, I found out the following key things:
1. "Culture" is used vaguely in Tanzanian policies, often to serve political agendas.
2. It's used to justify exclusionary laws on land, gender, and education.
3. Policy definitions of culture differ sharply from lived experiences.
4. Many accept policies as cultural truth due to institutional authority.
5. Heritage archives show a richer, more nuanced view of culture.
6. This gap reinforces marginalization and limits resource access.
Through this project, I aim to bring attention to culture as a tool that underserves some due to its vague nature that exists so prominently. To do so, I will invoke the use of Kitenge pattern regeneration systems that I will develop to create an interactive experience that explores this.
Critical analysis of culture
INSPIRATION
THE BIG IDEA
MID-FIDELITY PROTOTYPE
Building the system
REFLECTIONS AND NEXT STEPS
I’m building a generative, wall-mounted visualization system that transforms Tanzanian Kitenge prints into a dynamic storytelling interface—one that visually maps the discrepancy between government-defined and community-lived definitions of culture.
Using a combination of policy documents and cultural heritage archives—processed through large language models (LLMs)—the system extracts thematic definitions of culture and translates them into color and pattern data. These definitions are then rendered as evolving Kitenge-style prints in real time using p5.js, JavaScript, and interactive HTML interfaces.
The final installation will span across the walls of a space, where:
1. Fixed black boundaries represent rigid cultural narratives used in lawmaking.
2. Enclosed patterns and colors shift based on user-selected responses to cultural prompts, symbolizing how lived culture pushes against imposed definitions.
3. Each wall may represent a specific policy theme (e.g., land, gender, governance), turning the space into a layered, explorable data tapestry.
MID-FIDELITY PROTOTYPE
*Please interact with the screen below:
Critical analysis of culture
INSPIRATION
THE BIG IDEA
MID-FIDELITY PROTOTYPE
Building the system
REFLECTIONS AND NEXT STEPS
This prototype previews how the final Asili installation will function—transforming user input into evolving Kitenge prints that visualize the gap between institutional and lived definitions of culture.
Color Picker Input→ Represents participant definitions of culture through color.
Pattern Changes→ Simulates how lived culture disrupts static government narratives.
Black Boundaries→ Symbolize fixed, institutional definitions of culture from policy.
Visual Evolution→ Shows how collective input gradually reshapes the cultural narrative.
Theme Mapping→ Each cultural concept (e.g. tradition, governance) links to a specific print zone.
Critical analysis of culture
INSPIRATION
THE BIG IDEA
MID-FIDELITY PROTOTYPE
Building the system
REFLECTIONS AND NEXT STEPS
RECREATING THE KITENGE PRINT
Very little is known about what makes a print print a kitenge print. Significantly, I need to be able to make and alter patterns that resemble the kitenge patterns both in their initial and altered states (AI can't make kitenge prints) So I began to study the shapes, textures and colors of as many kitenge prints as I could process.
Here are my experiments:
Here are my experiments:
SHAPE STUDY
To understand the shapes that go into making a kitenge patterns, I used Node5.js to explore whether I could come even close to making a kitenge print shape with basic shapes
The results of my experiments were far from any shape I have seen on. a kitenge print. I learned that more organic and imperfect shapes were best for creating these prints. This realization pushed me to explore pattern and color recognition as a way forward.
COLOR STUDY
Using a custom python color extractor that i built, I was able to very accurately extract the colors present in my primary kitenge print. Studying the colors revealed that despite their vibrancy, these prints work best with more muted shades but that specific arrangements of these colors produced the vibrancy often seen in traditional prints.
Following the success while using my primary pattern that has more clearly focused shapes and patterns, I decided to experiment with a variation of the kitenge print that has a more faded/blurred design to see what lessons I could learn.
From these experiments, I learned the importance of texture in communicating a kitenge design/print. As such, I developed my own texture/noise to mimic the texture of the wax fabric that the kitenge prints are printed on. The results were as desired.
PATTERN STUDY
To further study patterns beyond trying to combine shapes, I coded a python system that detected patterns and then recombined them into organic shapes. Below is an early exploration where only a few patterns in my primary print were detected.
The results of recombining the select few patterns were beautiful new organic patterns that could morph into each other. Below are stills from a morphing experiment using only some patterns from my primary pattern.
Detecting all the patterns in my primary print combined with the colors I extracted and the texture layer I developed allowed me to use python to reconstruct my original pattern into a very close resembling print as seen in the middle.
WORKING ON INSTALLATION
This project is ultimately intended to live on the walls of a room-sized installation, where interactive Kitenge patterns respond to user input. I'm currently learning to use TouchDesigner, exploring real-time 3D geometry and camera mapping (as shown in the image) to prepare for large-scale spatial projection. This stage is focused on building the technical foundation that will allow me to bring the system into a physical space.
Critical analysis of culture
INSPIRATION
THE BIG IDEA
MID-FIDELITY PROTOTYPE
Building the system
REFLECTIONS AND NEXT STEPS
This project has been one of the most rewarding explorations I’ve undertaken—technically, conceptually, and personally. Recreating the vibrancy and logic of Kitenge prints has allowed me to engage deeply with cultural pattern-making as both a visual language and a political tool. It has also raised important questions for me about authorship, tradition, and ethical replication.
A key challenge I’ve grappled with is the ethics of using existing Kitenge prints as training material—extracting shapes and color patterns using Python to learn how they are constructed. While this was critical for understanding how to build a generative system, I remain conscious that these prints carry cultural and commercial value, and were created by designers and communities whose work should be credited and respected.
Going forward, I aim to use this foundational understanding not to copy but to create entirely new pattern systems inspired by the cultural logic of Kitenge, rather than the visual specifics of existing ones.
Technically, I’m still learning how to:
1. Use TouchDesigner to implement real-time morphing and pattern transitions on spatial surfaces (like walls).
2. Create patterns that respond dynamically to inputs over time, maintaining generative fluidity while preserving design integrity.
3. Balance visual complexity with performance across large projection surfaces.
Now that I’ve built a working pipeline and understand how to deconstruct and reconstruct Kitenge elements, my next steps are:
1. Design original generative Kitenge-style patterns from scratch.
2. Build a system that allows them to morph in real time in response to user interaction.
3. Expand from screen-based prototypes into a room-scale installation, where the walls will become responsive textiles.
This process has solidified my belief that code can be a form of cultural authorship—and that design systems can critique the systems that design us.
A key challenge I’ve grappled with is the ethics of using existing Kitenge prints as training material—extracting shapes and color patterns using Python to learn how they are constructed. While this was critical for understanding how to build a generative system, I remain conscious that these prints carry cultural and commercial value, and were created by designers and communities whose work should be credited and respected.
Going forward, I aim to use this foundational understanding not to copy but to create entirely new pattern systems inspired by the cultural logic of Kitenge, rather than the visual specifics of existing ones.
Technically, I’m still learning how to:
1. Use TouchDesigner to implement real-time morphing and pattern transitions on spatial surfaces (like walls).
2. Create patterns that respond dynamically to inputs over time, maintaining generative fluidity while preserving design integrity.
3. Balance visual complexity with performance across large projection surfaces.
Now that I’ve built a working pipeline and understand how to deconstruct and reconstruct Kitenge elements, my next steps are:
1. Design original generative Kitenge-style patterns from scratch.
2. Build a system that allows them to morph in real time in response to user interaction.
3. Expand from screen-based prototypes into a room-scale installation, where the walls will become responsive textiles.
This process has solidified my belief that code can be a form of cultural authorship—and that design systems can critique the systems that design us.
*Click to the report below to flip through it!
Overview
Investigating the opportunity areas for fitness furniture as a gym alternative in Berkeley, California.
Timeline
Ongoing (Prototype completed)
My Role
UX Researcher
Tools
FigJam, Airtable, NVivo, Dovetail, Statista, Mintel, Google Scholar, Notion.
Teammates (left to right)
Hongxi Pan
Sharon Zhao
Darlene Chen
THE DREAM TEAM!
Overview
A scooter add-on that disperses seeds while you move—bringing greenery to urban spaces and beyond.
Timeline
Ongoing (Prototype completed)
My Role
Product Designer & Fabrication Engineer
Tools
Autodesk Fusion 360, Arduino IDE, Figma, Ultimaker 3D Printer, Adafruit Feather M0 WiFi, P
Teammates
Elizabeth Sun
Sasha Suggs
Context
In this project, I explored how personal transportation can actively give back to the environment. Merging my fascination with sustainable tech and creative problem-solving, I co-designed a seed-dispersing device for scooters - planting seeds and transforming routine journeys into green adventures!
DEMO
Key Objectives
Research & Ideation
Process & Prototyping
Mobile App & Gamification
Final Design
REFLECTIONS AND NEXT STEPS
1. Reduce carbon emissions with a more sustainable personal mobility device.
2. Actively contribute to green spaces through seed dispersal.
3. Encourage positive behavior change via data and gamification.
Key Objectives
Research & Ideation
Process & Prototyping
Mobile App & Gamification
Final Design
REFLECTIONS AND NEXT STEPS
We began by asking ourselves:
“How can we make a scooter that benefits the environment while in motion?” This led to brainstorming features such as embedded seed capsules, fan-driven dispersion, and data tracking to measure impact.
Key Questions Explored:
“How can we make a scooter that benefits the environment while in motion?” This led to brainstorming features such as embedded seed capsules, fan-driven dispersion, and data tracking to measure impact.
Key Questions Explored:
1. How can we make a skateboard or scooter that actively benefits the environment?
2. What materials ensure sustainability without compromising functionality?
3. How can users track or visualize their positive impact in real time?
Key Objectives
Research & Ideation
Process & Prototyping
Mobile App & Gamification
Final Design
REFLECTIONS AND NEXT STEPS
Cardboard Prototyping
I used cardboard models for rapid prototyping. Cardboard allowed me to quickly test the form, function, and seed-dispersion mechanism with minimal cost. This phase helped me determine the optimal shape and size for portability and efficiency.
3D Printing & Fabrication
After testing with cardboard, I refined our design using 3D printing and biodegradable plastics. Key modifications included a compact form factor, improved material durability, and a refined fan-based seed dispersal mechanism. We iterated the design multiple times to ensure reliability and ease of use.
A key lesson in prototyping with 3D printing was learning how to balance flexibility (in anticipation of the abrasive use-case of riding a scooter) and rigidity to maintain the functionality of the key parts of its functional form. After experimenting and failing to find success with PLA-leaning materials, I looked into TPU leaning materials and the results were exactly what I was looking for.
%20(1).png)
Electronics & Real-Time Data Tracking
To bring our design to life, I integrated an electronics system that monitored seed dispersal in real time. I used the Arduino IDE along with the WiFi101 and Adafruit MQTT libraries to program an Adafruit Feather M0 WiFi microcontroller. I vividly recall the thrill of the first time the system logged data to the cloud—it wasn’t just numbers on a screen; it was the pulse of the project. This integration highlighted my journey from a creative designer to someone who could also harness tech to validate environmental impact.
Real-Time Data Tracking Setup:
Real-Time Data Tracking Setup:
1. Connect the Adafruit Feather M0 WiFi to a breadboard for power and control.
2. Mount a push-button on the breadboard: connect one terminal to GND and the other to a digital pin.
3. Using Arduino IDE with the WiFi101 and Adafruit MQTT libraries, write and upload code that starts/stops a timer and logs the seed dispersal event on Adafruit IO Cloud.
4. Watch as each button press creates a real-time data entry, confirming the action.
%20(4).png)
%20(5).png)
Key Objectives
Research & Ideation
Process & Prototyping
Mobile App & Gamification
Final Design
REFLECTIONS AND NEXT STEPS
Recognizing the importance of user experience, we developed a companion mobile app using Figma. The app isn’t just a dashboard—it’s a story of every seed planted. It tracks dispersal events, rewards users with points and badges, and suggests optimal planting routes based on real-time data. Every design decision here was personal, aiming to make environmental stewardship both engaging and visually rewarding.
FINAL DESIGN
STEP 1
Clip your planter on any ferromagnetic surface on your scooter.
Step 2
Pour your seeds into the nozzle that's release holes work best with the size of your seeds.
Step 3
Twist the top of that nozzle until you reach optimal seed flow.
STEP 4
Follow the map on your EcoPlanter app and enjoy planting!
REFLECTIONS AND NEXT STEPS
Working on Ride & Regrow taught me the importance of a methodical, iterative design process. I learned that each stage—from initial sketches and cardboard prototyping to 3D printing and electronics integration—provides essential data that refines the final product. I faced specific challenges, such as calibrating the fan-assisted seed dispersal system for consistent output and integrating reliable real-time data tracking with the Adafruit Feather M0. Balancing functionality with an appealing design required careful attention to both user feedback and technical constraints.
Next Steps:
Data Tracking Enhancement: Integrate GPS heatmaps and additional IoT sensors to improve the measurement and visualization of seed dispersion.
App Refinement: Upgrade the Figma prototype by incorporating machine learning algorithms for more accurate route suggestions and better gamification features.
Expand Collaborations: Initiate discussions with local government and sustainability organizations to explore practical applications and broaden the project’s impact.
I'm so happy to have worked with the team that Ididin developing this. Thank you to Elizabeth Sun for your detailed design insights and Sasha Suggs for ensuring precise data tracking and app integration.
Next Steps:
Data Tracking Enhancement: Integrate GPS heatmaps and additional IoT sensors to improve the measurement and visualization of seed dispersion.
App Refinement: Upgrade the Figma prototype by incorporating machine learning algorithms for more accurate route suggestions and better gamification features.
Expand Collaborations: Initiate discussions with local government and sustainability organizations to explore practical applications and broaden the project’s impact.
I'm so happy to have worked with the team that Ididin developing this. Thank you to Elizabeth Sun for your detailed design insights and Sasha Suggs for ensuring precise data tracking and app integration.
WHAT DID I BUILD?
I built a broom that turns your movement with it into real-time poetry and music. The broom analyzes how fast, hard, and rhythmically you move it, then uses sentiment analysis to create art that reflects the mood of your movement.
Timeline
Ongoing (Prototype completed)
My Role
Product Designer & Fabrication Engineer
Tools
Autodesk Fusion 360, Arduino IDE, Figma, Ultimaker 3D Printer, Adafruit Feather M0 WiFi, P
Teammates
Elizabeth Sun
Sasha Suggs
Why did I BUILD IT?
There are two main reasons I undertook this project:
1. Inspired by Don Norman’s The Design of Everyday Things and its critique that modern interfaces engage less with our bodies, I wanted to create more embodied interfaces that involve physical movement.
2. Reflecting on language accessibility, I realized that people who cannot communicate verbally often have to learn alternative languages (like ASL or fingerspelling). These methods can be limiting, as they require the audience to understand the same language.
These reflections led me to my primary question/exploration:
1. Inspired by Don Norman’s The Design of Everyday Things and its critique that modern interfaces engage less with our bodies, I wanted to create more embodied interfaces that involve physical movement.
2. Reflecting on language accessibility, I realized that people who cannot communicate verbally often have to learn alternative languages (like ASL or fingerspelling). These methods can be limiting, as they require the audience to understand the same language.
These reflections led me to my primary question/exploration:
What if language could be whatever the communicator desires, with technology translating it into familiar and accessible outputs?
Breaking this down
To explore this question, I built a broom device that, when used (and ideally danced with), generates music and text corresponding to the sentiments expressed by its movement.
So, why a broom?
So, why a broom?
My Process
1. 1. form - affrodances
To build this broom, I reimagined several parts of a standard design to house the electronics necessary for data collection. I identified the top of the broom handle as the ideal location since it was least likely to be interfered with by the user compared to the bottom.
1. 2. form - DESIGN
Designing the new broom top required considering a few key factors:
1. Ensuring the size comfortably accommodated all the electronic components without being visually overwhelming.
2. Allowing proper ventilation for these electronics.
Using Autodesk Fusion 360, TinkerCad, and a Prusa 3D printer, I created the enclosure shown on the left using sturdy PLA filament.
1. Ensuring the size comfortably accommodated all the electronic components without being visually overwhelming.
2. Allowing proper ventilation for these electronics.
Using Autodesk Fusion 360, TinkerCad, and a Prusa 3D printer, I created the enclosure shown on the left using sturdy PLA filament.
2.1. ELECTRONICS
The primary electronics and softwares for this project were:
1. MPU-6050 Accelerometer to collect motion data.
2. Arduino UNO R4 WiFi microcontroller to transmit location data to the cloud for processing.
3. The Python and Processing software stacks.
4. Custom location data retreival and processing API.
5. Arduino IDE
5. Supporting power and wiring.
1. MPU-6050 Accelerometer to collect motion data.
2. Arduino UNO R4 WiFi microcontroller to transmit location data to the cloud for processing.
3. The Python and Processing software stacks.
4. Custom location data retreival and processing API.
5. Arduino IDE
5. Supporting power and wiring.
3.1. Data collection
I initially wanted to learn how to analyze movement data. My idea was to begin by mapping specific types of motion to corresponding sentiments in a controlled environment. To achieve this, I collaborated with artist and choreographer River James to create choreography for analysis.
3.2 DEFINING THE MOTION
After gathering location data from the choreography, I developed a Python algorithm to sort and categorize the repeated motions. I systematically analyzed motion clips and fine-tuned the algorithm's sensitivity to ensure that grouped motions were similar and intentionally reproduced by the dancer.
These grouped patterns were then mapped to specific emotional cues to inform the system's eventual writing. On the right, you can see the evolution of my analysis and fine-tuning process, with each colored line representing a different motion pattern.
These grouped patterns were then mapped to specific emotional cues to inform the system's eventual writing. On the right, you can see the evolution of my analysis and fine-tuning process, with each colored line representing a different motion pattern.
3.3 RETRACING MOTION
After clustering the data from the choreographer, I focused on guiding someone to recreate the motion. With movements mapped in x, y, and z directions, repeating the motion was challenging. I developed a feedback system using dots and a sphere to show your position relative to the detected range. As you move, you receive real-time feedback on the closest predefined cluster and guidance on how to adjust. Some experiments are shown on the left including an early experiemnets with using flowfield visualizations to accomplish this.
4.1 . ADDING THE TEXT
For this iteration, I decided to use a limited dataset so I could more closely examine the system's performance. I used this dataset to train my Python model through sentiment analysis, subsequently mapping the resulting language groupings and syntax patterns to the various movement clusters I had collected. The dataset comprised a collection of African poetry, much of which can be found in "Modern Poetry from Africa."
4.2 . SOUND
To provide feedback on what the user is communicating, I added a sound element to the system. This audio cue indicates how close a user is getting to a specific emotional cluster, with a brief delay that allows the user to adjust their movements if the output doesn’t match their intended emotional expression.
The result is an instrument of music and text-based communication.
The result is an instrument of music and text-based communication.
FINAL RESULTS, reflections and next steps
The outcome is a broom that can be used as a musical instrument, but one that can also be programmed to translate your movements into written language. More immediately, the broom now produces music and poetry when it is danced with.
This project was not only fun but also an enlightening opportunity to learn about the limitations in communication faced by disabled individuals through participant interviews. I learned that inclusive technology does not isolate its users; rather, it works to provide them with access and presence in the shared spaces around us.
While the project is a successful prototype, I recognize the need for more efficient sensors to handle the real-time data transfers and processing taking place. I also envision this project evolving into a wearable product featuring a more streamlined motion cluster training system and a comprehensive language-to-cluster mapping, which would allow for even more detailed communication.
This project was not only fun but also an enlightening opportunity to learn about the limitations in communication faced by disabled individuals through participant interviews. I learned that inclusive technology does not isolate its users; rather, it works to provide them with access and presence in the shared spaces around us.
While the project is a successful prototype, I recognize the need for more efficient sensors to handle the real-time data transfers and processing taking place. I also envision this project evolving into a wearable product featuring a more streamlined motion cluster training system and a comprehensive language-to-cluster mapping, which would allow for even more detailed communication.
Let's Connect!
