AI Drive-thru voice ordering
An AI product built for restaurant drive-through's to help customers order their food.
Case studies
GreenTogether
A three member team project focused on sustainability and helping users find and participate in local environmental events such as park cleanups, recycling drives, and tree planting.
Graphic design
A comprehensive view of my work as an artist.
AI Drive-thru voice ordering
Overview
Problem
Staffing shortages and human error make drive-thru ordering inconsistent — and an AI taking the order only helps if the customer believes it got the order right.
Goal
Make an order-taker people are willing to use: one a customer can check without slowing the lane down.
My role
User research, competitive research, journey mapping, UX, UI.
Solution
I show the order forming as it is spoken — a live transcript, a visible listening state, and menu items labelled with the exact words that order them — so a mistake gets caught before the window, not at it.
Result
Built for fast-food chains, delivered in two weeks.
Key features — Drive-thru
Voice-controlled menu
Every item is labeled with the exact words that order it — “Add Flame Master” — so no one has to guess the phrasing the AI accepts. We also designed it for any type of device and size.
The order panel fills in beside the menu as items land, keeping the running total in view before anyone commits to it.
Conversation history
The transcript stays on screen so customers can check what the AI heard against what they actually said. When something’s wrong, they can see exactly where it went wrong.
Before vs After
Before
After
AI Host
The AI Host was created to help the users feel more comfortable communicating with the AI. The AI Host visuals and voice are entirely customizable for every restaurant.
A “Listening…” notification lights up whenever the mic is open, so people aren’t left talking into a screen that might not be paying attention. The host itself — avatar and voice — is customizable per restaurant.
What research told me
- Trust was the blocker — not speed, and not accuracy on paper.
- People didn’t distrust the AI for being AI. They distrusted it because they couldn’t tell whether it had heard them.
- The fix was feedback, not a smarter model. Show the order forming while it’s being spoken.
Designing the failure cases
- When the model isn’t confident it asks — “Diet Coke or regular?” — instead of guessing.
- Every item in the running order stays editable, so a correction is one tap, not a restart.
- Misheard input, background noise and mid-order changes each got their own flow.
How I worked
- Competitive review first, then the happy path — then I mapped the unhappy path, which is where nearly every feature here came from.
- Wireframes went straight into moderated tests: five people ordering out loud to a screen.
- I cut the AI Host’s spoken confirmations after testing — hearing every item read back made the wait feel longer, so the reassurance moved to the visual listening state.
What I’d change
- Bring in real background noise earlier. I ran early sessions in a quiet room, the one condition a drive-thru is never in.
- Test the AI Host’s personality against a plain progress indicator — I never proved the character earns its place over the transparency features.
GreenTogether
Project overview
Problem
More than 8 million pieces of trash a year are dropped outside bins in public places — and picking some up alone feels pointless, so people do it once and stop.
Goal
Turn a one-off good deed into a habit: give people proof the effort adds up, and a reason to come back next weekend.
My role
Research, UX, UI — a 24-hour design competition with a team of four.
Solution
I paired a weekly-resetting leaderboard, so a newcomer is never permanently behind, with a map that turns the intention into a specific event to show up to.
Result
Judges called the map and leaderboard pairing the clearest read on the problem of any team that day.
Key features
Leaderboard
The leaderboard makes the effort add up to something visible — a running count that gives people a reason to come back.
Interactive map
Once the leaderboard gave people a reason to show up, they needed a way to find something to show up to. The map shows the locations of the events, with dates and times below.
What research told me
- Awareness wasn’t the blocker — people already knew littering was a problem.
- Picking up trash alone felt pointless. They wanted proof the effort added up.
- And they wanted a reason to come back the next weekend.
How I worked
- Three concepts sketched in the first two hours, voted on as a team before anyone opened Figma — cheaper to kill an idea on paper.
- The leaderboard came first; the map came out of a later pivot, once we saw people needed an event to show up to, not just a score.
- Our first leaderboard ranked by lifetime total — unwinnable for anyone starting late. Weekly resets put everyone on the same clock each Monday.
What I’d change
- We spent the 24 hours on research and flows, and the visual execution never caught up — against the finalists, that showed.
- Next time: lock the flows by hour twelve and give the back half to visual design.
Graphic design
CalState Fullerton Discoverfest logo
Version 1
The client hadn’t settled on a theme for Discoverfest and left that to us, so round one was really a set of theme proposals. The brief asked for a text-based mark, so I built the concept into the type itself — a globe in place of the O, pushing exploration and discovery without adding a separate illustration.
Version 2
The first concept, “Puddle Splash,” didn’t survive review — the splash shape and droplets competed with the type, so the thing people were meant to read came second. “Building Blocks” put the idea into the letterforms instead of around them: “Discover” and “Fest” on tilted 3D blocks, students building their own Titan Experience.
Version 3 — final
The client picked Building Blocks. I built out vertical and horizontal lockups so the mark works across banners, posters and social.
Education
2026–2027
University of California, Irvine
Master of Human-Computer Interaction & Design. Graduating August 2027.
2024–2026
California State University, Fullerton
Bachelor of Fine Arts (BFA) in Graphic and Interactive Design. Graduated with Honors.
2022–2024
Saddleback College
Associate in Arts for Transfer, Studio Arts, with an emphasis in Digital Arts and Social Psychology. Graduated with Honors.