T-Mobille
CareLink

Practices

AI Tools, Communication

Tools

Cursor, Colab, Firebase

T-Mobile aims to address the challenge of monitoring ACL recovery progress at-home by designing a tool for real-time monitoring and remote doctor support.

Role

Product Designer

Time

Mar - Jun.2025

Practices

AI Tools, Communication

Tools

Cursor, Firebase, Colab

Practices

AI Tools, Communication,

Hardware & Software

What Makes This Project Unique？

There are three key aspects that make our project stand out.

There are three key aspects that

make our project stand out:

01 AI Tools Applied

Cursor is the AI tool we use most frequently, along with ChatGPT and Claude. After completing this project, I divided my AI usage experiences into two areas-coding and design. Below are my thoughts on these tools:

After completing this project, I divided my AI usage experiences into two areas-coding and design. Below are my thoughts on these tools:

Part 1#

Coding

Cursor

The main platform for building coding projects.

👍 1. Build both front-end and back-end frameworks

2. View and interact with various project files.

3. Auto mode selects suitable AI models.

👎 Occasionally deletes files, requiring careful validation before

accepting Cursor’s solution.

Using Cursor for Coding

The main platform for building coding projects.

👍 1. Build both front-end and back-end frameworks

2. View and interact with various project files.

3. Auto mode selects suitable AI models.

👎 Occasionally deletes files, requiring careful validation before

accepting Cursor’s solution.

ChatGPT

Address logical and method issues.

👍 1. Strong understanding.

2. Provide good project logic suggestions.

👎 Cannot directly access or analyze the file structure,

requiring manual input, which can be time-consuming.

Claude

Mainly used to debug code.

👍 1. Excellent at debugging code.

2. Directly displays the corresponding visual interface for the

code, allowing interaction to check the results.

👎 Although it provides an interactive interface, it cannot execute

actual deployment.

Part 2#

Design

Using MCP and code, Cursor can communicate directly with Figma.

👍 Can transform Figma designs into code.

👎 Although Cursor can convert Figma designs into code directly,

the generated code often differs significantly from the original

design and requires extensive manual adjustments to be usable.

Just a prompt, Claude can generate an editable SVG file for Figma.

👍 Useful for brainstorming design ideas.

👎 The SVG files generated are not fully polished and require

significant manual adjustments.

Part 1#

Coding

Cursor

The main platform for

building coding projects.

👍 1. Build front-end and back-end works.

2. Interact with various project files.

3. Auto selects suitable AI models.

👎 Sometimes deletes files, requiring validation.

Using Cursor for Coding

ChatGPT

Address logical and method issues.

👍 1. Strong understanding.

2. Provide project logic suggestions.

👎 Cannot directly access or analyze the file structure, requiring manual input.

Claude

Mainly used to debug code.

👍 1. Excellent at debugging code.

2. Directly displays the corresponding visual interface for the code, allowing interaction to check the results.

👎 Though it provides an interactive interface, it cannot execute

actual deployment.

Part 2#

Design

Using MCP and code, Cursor can communicate directly with Figma.

👍 Can transform Figma designs into code.

👎 Code output may not match the original design and often needs further refinement.

Just a prompt, Claude can generate an editable SVG file for Figma.

👍 Useful for brainstorming design ideas.

👎 The SVG files generated are not polished and require manual adjustments.

01 AI Tools Applied

Part 1#

Coding

Cursor

The main platform for building coding projects.

👍 1. Build both front-end and back-end frameworks

2. View and interact with various project files.

3. Auto mode selects suitable AI models.

👎 Occasionally deletes files, requiring careful validation

before accepting Cursor’s solution.

Using Cursor for Coding

ChatGPT

Address logical and method issues.

👍 1. Strong understanding.

2. Provide good project logic suggestions.

👎 Cannot directly access or analyze the file structure,

requiring manual input, which can waste time.

Claude

Mainly used to debug code.

👍 1. Excellent at debugging code.

2. Directly displays the corresponding visual

interface for the code, allowing interaction to

check the results.

👎 Although it provides an interactive interface, it

cannot execute actual deployment.

Part 2#

Design

Using MCP and code, Cursor can communicate

directly with Figma.

👍 Can transform Figma designs into code.

👎 Although Cursor can convert Figma designs into

code directly, the generated code often differs

significantly from the original design and requires

extensive manual adjustments to be usable.

Just a prompt, Claude can generate an editable

SVG file for Figma.

👍 Useful for brainstorming design ideas.

👎 The SVG files generated are not fully polished

and require significant manual adjustments.

02 Stakeholder Communication

We closely collaborated with our sponsor T-Mobile, ACL recovery patients, and our instructors, which significantly enhanced our communication skills. Meanwhile, we received the guidance from UW. Below is our instructor team.

We collaborated with T-Mobile, ACL patients, and UW instructors, which greatly improved our communication skills.

Candice Boyd

Senior Manager, T-Mobile

Candice Boyd

Senior Manager,

T-Mobile

John Raiti

Associate Teaching Professor, UW

Luyao Niu

Assistant Teaching Professor, UW

Luyao Niu

Assistant Teaching Professor,

Candice Boyd

Senior Manager, T-Mobile

STEP 1 Initial Discussion:

STEP 1 Initial Discussion:

Concept Development

🤝 Communication key point:

Communicated with T-Mobile via email and virtual meetings to confirm project direction
Introduced team members and shared early ideas
A clear understanding of sponsor expectations to guide concept development

🤝 Communication key point:

Synced with T-Mobile via email and virtual meetings to align on direction
Introduced team & shared early ideas
Clarified sponsor expectations to guide concept development

Email and online meeting

Discussion offline

Email and online meeting

Discussion offline

STEP 2 Millstone:

STEP 2 Millstone:

T-Mobile Visit and Project Update

🤝 Communication key point:

Shared research insights and design progress during the on-site visit
Collected valuable feedback from the sponsor to refine our direction

Group Photo with Sponsor

T-Mobile 5G Innovation Hub

STEP 3 Final:

STEP 3 Final:

Presentation and Real-Time Demo

🤝 Communication key point:

Presented our final solution and conducted a live demo
Explained our design decisions and responded to sponsor feedback
Discussed about future direction and next step

🤝 Communication key point:

Presented and conducted a live demo
Explained our design decisions and responded to sponsor feedback
Discussed about future direction

STEP 2 Millstone:

T-Mobile Visit and Project Update

🤝 Communication key point:

Shared research insights and design progress during the on-site visit
Collected valuable feedback from the sponsor to refine our direction

Group Photo with Sponsor

T-Mobile 5G Innovation Hub

I'm presenting!

Group Photo with T-Mobile Sponsor

I'm presenting!

Group Photo with T-Mobile Sponsor

03 Hardware & Software

We finished the project from brainstorming, designing, hardware setting and software connection. Here I showed some brief photos, if you're interested in the process, please send me an email, I will share more details with you :)

👭 Meet us#

Our Team

💡

Step 1#

Brainstorm

We iteratively developed a wearable knee brace with embedded IMU sensors.

DIscussion

Whiteboard Notes

Step 2#

Hardware

Design

Evolved from a bulky mechanical brace to a lightweight, sensor-integrated wearable

optimized for user comfort and mobility during ACL rehabilitation.

1st Design

2nd Design

3rd Design

Step 3#

Machine
Learning

A Random Forest Regressor was used to recommend personalized rehabilitation angle goals

based on each individual's specific condition.

Training Data Collection

Machine Learning Result

Step 4#

Software
Design

Live data from the knee is transmitted via microcontroller to a Firebase backend, then visualized on a web-based dashboard. The website was built with React and deployed online.

Patient Dashboard

Doctor Dashboard

👭 Meet us#

Our Team

03 Hardware & Software

💡

Step 1#

Brainstorm

We developed a wearable knee brace with embedded IMU sensors.

Discussion

Whiteboard Notes

Step 2#

Hardware

Design

Evolved from a bulky mechanical brace to a lightweight, sensor-

integrated wearable optimized for user comfort during ACL rehabilitation.

1st Design

2nd Design

3rd Design

Step 3#

Machine
Learning

A Random Forest Regressor was used to recommend personalized

rehabilitation angle goals based on each individual's specific condition.

Training Data Collection

Machine Learning Result

Step 4#

Software
Design

Live data from the knee is transmitted via microcontroller to a Firebase backend, then visualized on a web-based dashboard. The website was built with React and deployed online.