Conducting senior thesis research in vision + NLP, specifically methods for unsupervised learning of image captioning.
Built back-end infrastructure and ML models on Machine Intelligence team of Google Photos.
Created Pixplor, an "intelligent photo album" iPad app for nursing-home residents with dementia. Check out this article about the project.
We introduce a perceptually motivated approach to bandwidth expansion for speech. Our method pairs a new 3-way split variant of the FFTNet neural vocoder structure with a perceptual loss function, combining objectives from both the time and frequency domains. Mean opinion score tests show that it outperforms baseline methods from both domains, even for extreme bandwidth expansion.
Collaborators: Zeyu Jin, Jiaqi Su, Prof. Adam Finkelstein
This project began as an Independent Work project and won the Princeton CS Department's Best Poster Award. It developed into a full-fledged research project and publication.
Analyzing the effect of different text encodings on visual question answering (VQA) models.
Tau Beta Pi
Lyman Prize (Valedictorian)
State Champion and NSDA National Quarterfinalist in Extemporaneous Speaking
I love music and believe good music comes from all genres, instruments, and people. I produce music using a mix of digital/analog synthesis (feel free to check out my SoundCloud). I play piano (my current goal is to master Chopin's Ballade No. 1 in g minor) and am teaching myself guitar.
When not geeking out over machine learning or music theory, I enjoy playing tennis and watching reruns of Friends.