Andi's selfie. I'm an Asian woman with long hair. I dyed part of my hair white-gold. I'm wearing a blue cap and a gray hoodie.

I’m currently a master student in Stanford University majoring in Computer Science. I earned a BS in Computer Science at University of Michigan, Ann Arbor, with minors in Mathematics and Japanese.

I’m interested in using AI to augment human capabilities and am motivated by the question, “How can we build systems that support humans while still ensuring their agency in this new AI era?”

In the past, I worked on HCI, accessibility and Human-AI interaction, fortunate to work under the supervision of Prof. Anhong Guo at Human-AI Lab.

News

Sep 2024: Started my master at Stanford. May 2024: Started my internship at Meta. Dec 2023: Nominated as a runner-up for CRA Outstanding Undergraduate Researcher Awards.
Oct 2023: Attend ASSETS 2023 in NYC.

Research
Many flat interfaces are inaccessible to blind and visually-impaired people. This photo shows a person using the microwave panel with the help of our Vizlens app.
Deploying VizLens: Characterizing User Needs, Preferences, and Challenges of Physical Interfaces Usage in the Wild
Andi Xu, Mahdi Qazwini, Chen Liang, Anhong Guo
ASSETS 2023 Demo
pdf  ·  ACM DL  ·  full video  ·  demo video
AI-based assistive technologies tend to assume 'universal' needs of BVI people and are thus one-size-fits-all, rather than accounting for unique differences and desires. This figure shows an example where a blind user uses Seeing AI to read a letter and trying to find out about the sender name Rebecca. However, it'll read many other unnecessary and distracting details.
Hacking, Switching, Combining: Understanding and Supporting DIY Assistive Technology Design by Blind People
Jaylin Herskovitz, Andi Xu, Rahaf Alharbi, Anhong Guo
CHI 2023
pdf  ·  ACM DL  ·  full video  ·  30s preview  ·  talk  ·  dataset
Project
A screenshot of our game. The background is a planet in the universe. Punching a key on the virtual keybaord emits a sparkling magic to hit a coming comet.
Website Code

Theseus Orbital Station

A VR typing game to study optimal typing experiences in virtual reality.

A photo of a hand wearing our ring hardware. Its parts span over index finger and wrist.
Video Report

One ring to rule them all

A ring to improve typing and clicking experience in virtual reality.

A diagram of our model. We extracted 16 frames from a video, process it with a spectral feature extractor, put it into a classifier, and use majority vote at the end.
Report

DeepFake Video Detection

Machine Learning course final project. Unimodal methods for DeepFake video classification with spectral features.

Diagram of our model. We used a Twitter sentiment analysis module and a stock price time series analysis module to together predict stock price.
Report

Stock Price Prediction

Affective computing course final project. Twitter (not X at that time) sentiment analysis and time series analysis.