Details

Role: UI/UX Designer
Timeline: Fall 2025 – Spring 2026
Team: ScottyLabs AI Team
Status: Live in beta
Tools: Figma

Summary

Bark (formerly CMUGPT) is an AI assistant for the Carnegie Mellon community, built with ScottyLabs. It pulls information from other ScottyLabs platforms like CMUEats, CMU Courses and CMU Maps, so students can ask questions about CMU in their own words and get answers specific to their needs.

An iMac showing the live Bark answering how to get from Tepper to Margaret Morrison with an interactive CMU Maps route

Overview

Problem

CMU students struggle to find the campus information they need

Finding answers to everyday questions about CMU often takes more effort than it should.

Solution

An AI assistant built for CMU students

Bark lets students ask questions about CMU in their own words and get answers specific to their needs, all in one chat.

Problem

Campus info is scattered

Dining, directions, course info, and other campus resources each live on a different site, so students have to guess where to look before they can get an answer.

Solution

One place to ask campus questions

Students can ask about dining, directions, course planning, and more without figuring out which website to visit first.

Problem

General AI may lack current CMU information

General-purpose chatbots may not have the latest dining hours, course details, or campus directions.

Solution

Answers use current campus data

Bark retrieves information from ScottyLabs platforms and links students to the relevant resource.

Problem

Useful answers are hard to find again

Once students find an answer, they often have to search for it all over again the next time.

Solution

Search and star past chats

Students can search their chats and star the ones they use most to find useful information again.

Research

Student interviews

Understanding how students find campus information

I interviewed CMU students to understand how they find campus information, which resources they use, and where they encounter difficulties.

Sample
8 participants
Timeline
1–2 weeks
Method
Semi-structured, one-on-one interviews

In students’ own words:

“There are so many CMU websites, and I don’t really know what information is on each one.”
CMU sophomore
Mechanical Engineering major
“I can Google what’s open on campus, but I still have to check if those hours are right for today.”
CMU junior
Information Systems major

Key findings

01

Campus information is scattered and difficult to navigate

Campus information is spread across different websites. Students often aren’t sure which site has what they need, so they have to search several sites for an answer.

02

General search tools lack campus-specific context

Students’ questions about CMU often require up-to-date campus information that general search engines and AI assistants may not provide.

Design opportunity

How might we help CMU students find campus information without searching across multiple websites?

Design principles

I used the research findings and discussions with the ScottyLabs team to define two principles for the design.

01

Keep search quick and easy

Let students start typing right away, without extra clicks or distractions.

02

Follow the ScottyLabs design system

Use ScottyLabs’ existing components and visual styles to keep the experience consistent with its other platforms, while giving Bark its own personality.

Process

Iterations

From initial prototype to v1

My initial prototype followed the layout of familiar AI chat tools. Through a review with the ScottyLabs team and user testing, I identified three key areas for improvement.

Sample
5 CMU students
Testing period
1 week
Method
Moderated usability testing
Initial prototype home screen: a plain Hi there! greeting and message box, with recent chats in a grey sidebar
Home
Initial prototype directions answer with four vague text steps and a static map card
Directions answer

Student quotes

“It looks like ChatGPT. I’m not really sure what’s different about it.”
CMU Sophomore · Information Systems
“I wish I could just see the route on the map instead of reading all these steps.”
CMU Junior · Mechanical Engineering

Key insights

01

Bark’s purpose wasn’t immediately clear

The initial interface felt similar to other AI chatbots and didn’t communicate what made Bark useful for CMU students.

02

Students needed more visual, engaging answers

Students found long text responses less engaging, and text-based instructions were difficult to follow, especially for campus directions.

03

Students struggled to find previous answers

As their chat history grew, students had difficulty finding specific answers from past conversations.

What didn’t work in the initial prototype:

Looked like any other chatbot

The interface looked like a generic chatbot and didn’t reflect the ScottyLabs brand system.

Directions were hard to follow

Text steps like “find your way toward the path” were vague, and the map was only a static preview.

Hard to get back to answers

There was no way to save useful answers to easily return to them later.

Key design shifts

Generic chatbot → CMU-specific assistant

For v1, I rebuilt the interface with ScottyLabs’ components and colors. Students sign in with their CMU account, and Bark greets them by name.

Live Bark home screen with the Bark logo, a light blue sidebar with Starred and Recents, the greeting Hi June, where are we headed on campus?, and the signed-in CMU account Bark logo and ScottyLabs colorsGreets each student by nameSigned in with a CMU account

Text answers → live ScottyLabs tools

Instead of long text answers, Bark now pulls live information from ScottyLabs tools like CMU Maps, CMUEats and CMU Courses. For directions, the vague text steps became an interactive CMU Maps route right in the chat.

Initial prototype answering how to get to Margaret Morrison from Tepper with four vague text steps and a Get directions on CMU Maps link
Before: initial prototype
Live Bark directions answer with an interactive CMU Maps route, arrival time, distance and route options
After: live v1

One-off chats → saved chats and memory

Students can now star a chat to keep it at the top of the sidebar, and Bark saves useful details to memory for later chats.

Live Bark chat showing a Memory saved note and a star button in the header Star a chat to keep itMemory saved for later chats

Final design

Key features

01

A simple home screen

Start typing right away.

After signing in with their CMU account, students can ask a question, choose an AI model, or reopen a recent chat from the sidebar.

02

ScottyLabs tools in the chat

Find campus information without leaving Bark.

Bark pulls information from CMU Maps, CMUEats, and CMU Courses to answer questions about directions, dining, and classes.

When students ask for directions, Bark shows an interactive map with the route distance, estimated arrival time, and options like Accessible and Inside. Students can also open the full route in CMU Maps.

03

Search and starred chats

Find an old answer without scrolling.

Search shows matching chats as cards, each with its title and when it was last used.

Students can star a chat from its header to keep it in the Starred list at the top of the sidebar.

04

Memories

Answers that already know you.

When a student shares something about themselves, like their school or major, Bark saves it as a memory and shows a small Memory saved note in the chat.

Memories carry into future chats so answers fit the student. They can review or delete each one from Memories in the account menu.

Next steps

Future developments

What’s next for Bark

I’m currently working with the ScottyLabs AI team to bring these updates to Bark, based on usage data and feedback from the first version. These updates focus on making Bark more intuitive and useful in students’ everyday campus lives.

Personalized prompt cards

The home screen will suggest questions on four prompt cards with topic icons and a Shuffle button.

Next-iteration home screen with four prompt cards, a Shuffle button and the message box Four suggested promptsShuffle for a new setOr type any question

The cards pull from four signals, in order of priority:

01

Personal history

Follows up on topics a student asked about before, like finishing a schedule.

02

Semester stage

Matches the academic calendar, from finding clubs in week one to graduation checks.

03

Time and context

Uses time of day and day of week, so lunch questions show up around noon.

04

Trending

Fills any remaining gaps with what other students across campus are asking about this week.

Pinned chats and search previews

A Pin page will group saved chats by month, and search results will preview each chat.

Design screen: search results for Plan schedule 2026 spring shown as chat cards with previews
Search with previews
Design screen: pinned chats grouped by This month, March and February
Pin page

Making the wait feel more engaging

I’m exploring two loading animations to show when Bark is gathering information: a running Scotty dog and a blue dot with a ripple effect.

Design screen: a chat showing two Fetching loading states while Bark gathers campus data, one with a running Scotty dog and one with a blue dot Option 1: a running Scotty dogOption 2: a blue dot with a ripple effect

Reflection

What I learned

01

Designing alongside developers

Working closely with the ScottyLabs AI team helped me understand how an AI product moves from design to launch. I learned to explain my design decisions, discuss feasibility with developers, and work with them to refine the interface.

02

Designing around student needs

Talking directly with students helped me understand their frustrations with existing campus resources and identify what Bark needed to do differently. I learned to turn research findings and testing feedback into design decisions that addressed real student needs.